Search Results

Search found 611 results on 25 pages for 'bare'.

Page 21/25 | < Previous Page | 17 18 19 20 21 22 23 24 25  | Next Page >

  • How to use DI and DI containers

    - by Pinetree
    I am building a small PHP mvc framework (yes, yet another one), mostly for learning purposes, and I am trying to do it the right way, so I'd like to use a DI container, but I am not asking which one to use but rather how to use one. Without going into too much detail, the mvc is divided into modules which have controllers which render views for actions. This is how a request is processed: a Main object instantiates a Request object, and a Router, and injects the Request into the Router to figure out which module was called. then it instantiates the Module object and sends the Request to that the Module creates a ModuleRouter and sends the Request to figure out the controller and action it then creates the Controller and the ViewRenderer, and injects the ViewRenderer into the Controller (so that the controller can send data to the view) the ViewRenderer needs to know which module, controller and action were called to figure out the path to the view scripts, so the Module has to figure out this and inject it to the ViewRenderer the Module then calls the action method on the controller and calls the render method on the ViewRenderer For now, I do not have any DI container set up, but what I do have are a bunch of initX() methods that create the required component if it is not already there. For instance, the Module has the initViewRenderer() method. These init methods get called right before that component is needed, not before, and if the component was already set it will not initialize it. This allows for the components to be switched, but it does not require manually setting them if they are not there. Now, I'd like to do this by implementing a DI container, but still keep the manual configuration to a bare minimum, so if the directory structure and naming convention is followed, everything should work, without even touching the config. If I use the DI container, do I then inject it into everything (the container would inject itself when creating a component), so that other components can use it? When do I register components with the DI? Can a component register other components with the DI during run-time? Do I create a 'common' config and use that? How do I then figure out on the fly which components I need and how they need to be set up? If Main uses Router which uses Request, Main then needs to use the container to get Module (or does the module need to be found and set beforehand? How?) Module uses Router but needs to figure out the settings for the ViewRenderer and the Controller on the fly, not in advance, so my DI container can't be setting those on the Module before the module figures out the controller and action... What if the controller needs some other service? Do I inject the container into every controller? If I start doing that, I might just inject it into everything... Basically I am looking for the best practices when dealing with stuff like this. I know what DI is and what DI containers do, but I am looking for guidance to using them in real life, and not some isolated examples on the net. Sorry for the lengthy post and many thanks in advance.

    Read the article

  • When things go awry

    - by Phil Factor
    The moment the Entrepreneur opened his mouth on prime-time national TV, spelled out the URL and waxed big on how exciting ‘his’ new website was, I knew I was in for a busy night. I’d designed and built it. All at once, half a million people tried to log into the website. Although all my stress-testing paid off, I have to admit that the network locked up tight long before there was any danger of a database or website problem. Soon afterwards, the Entrepreneur and the Big Boss were there in the autopsy meeting. We picked through all our systems in detail to see how they’d borne the unexpected strain. Mercifully, in view of the sour mood of the Big Boss, it turned out that the only thing we could have done better was buy a bigger pipe to and from the internet. We’d specified that ‘big pipe’ when designing the system. The Big Boss had then railed at the cost and so we’d subsequently compromised. I felt that my design decisions were vindicated. The Big Boss brooded for a while. Then he made the significant comment: “What really ****** me off is the fact that, for ten minutes, we couldn’t take people’s money.” At that point I stopped feeling smug. Had the internet connection been better, the system would have reached its limit and failed rather precipitously, and that wasn’t what he wanted. Then it occurred to me that what had gummed up the connection was all those images on the site, that had made it so impressive for the visitors. If there had been a way to automatically pare down the site to the bare essentials under stress… Hmm. I began to consider disaster-recovery in the broadest sense – maintaining a service in spite of unusual or unexpected events. What he said makes a lot of sense: sacrifice whatever isn’t essential to keep the core service running when we approach the capacity limits. Maybe in IT we should borrow (or revive) the business concept of the ‘Skeleton service’, maintaining only the priority parts under stress, using a process that is well-prepared and carefully rehearsed. How might this work? Whatever the event we have to prepare for, it is all about understanding the priorities; knowing what one can dispense with when the going gets tough. In the event of database disaster, it’s much faster to deploy a skeletal system with only the essential data than to restore the entire system, though there would have to be a reconciliation process to update the revived database retrospectively, once the emergency was over. It isn’t just the database that could be designed for resilience. One could prepare for unusually high traffic in a website by designing a system that degraded gradually to a ‘skeletal’ site, one that maintained the commercial essentials without fat images, JavaScript libraries and razzmatazz. This is all what the Big Boss scathingly called ‘a mere technicality’. It seems to me that what is needed first is a culture of application and database design which acknowledges that we live in a very imperfect world, and react accordingly when things go awry.

    Read the article

  • HTML, JavaScript, and CSS in a NetBeans Platform Application

    - by Geertjan
    I broke down the code I used yesterday, to its absolute bare minimum, and then realized I'm not using HTML 5 at all: <html> <head> <link rel="stylesheet" href="style.css" type="text/css" media="all" /> <script type="text/javascript" src="https://ajax.googleapis.com/ajax/libs/jquery/1.6.3/jquery.min.js"></script> <script type="text/javascript" src="https://ajax.googleapis.com/ajax/libs/jqueryui/1.8.16/jquery-ui.min.js"></script> <script type="text/javascript" src="script.js"></script> </head> <body> <div id="logo"> </div> <div id="infobox"> <h2 id="statustext"/> </div> </body> </html> Here's the script.js file referred to above: $(function(){ var banana = $("#logo"); var statustext = $("#statustext"); var defaulttxt = "Drag the banana!"; var dragtxt = "Dragging the banana!"; statustext.text(defaulttxt); banana.draggable({ drag: function(event, ui){ statustext.text(dragtxt); }, stop: function(event, ui){ statustext.text(defaulttxt); } }); }); And here's the stylesheet: body { background:#3B4D61 repeat 0 0; margin:0; padding:0; } h2 { color:#D1D8DF; display:block; font:bold 15px/10px Tahoma, Helvetica, Arial, Sans-Serif; text-align:center; } #infobox { position:absolute; width:300px; bottom:20px; left:50%; margin-left:-150px; padding:0 20px; background:rgba(0,0,0,0.5); -webkit-border-radius:15px; -moz-border-radius:15px; border-radius:15px; z-index:999; } #logo { position:absolute; width:450px; height:150px; top:40%; left: 30%; background:url(bananas.png) no-repeat 0 0; cursor:move; z-index:700; } However, I've replaced the content of the HTML file with a few of the samples from here, without any problem; in other words, if the HTML 5 canvas were to be needed, it could seamlessly be incorporated into my NetBeans Platform application: https://developer.mozilla.org/en/Canvas_tutorial/Basic_usage

    Read the article

  • Where to find new Micro-BTX (uBTX) motherboards? Or should I just replace the box?

    - by John Rudy
    OK, so I'm guessing that it's dead. It's not my machine, and the owner is on a very fixed (IE, none) income. I'm generous, but I'm not that generous, since I already gave him what (at the time) was a fully functional and fairly well-equipped machine. (Aside from the mobo and proc, almost nothing else in it was stock. I'd taken it up to 3GB of RAM, upgraded the hard drive, added a decent video card, installed a wireless adapter, running Vista, etc.) According to further research, the machine uses a Micro-BTX (uBTX) motherboard, and since it's an AMD Athlon64, the AM2 socket. So I'm looking at a few options, and am wondering what's the best route to take? Find an AM2 socket uBTX mobo. I can't find them new online anywhere, leading me to believe that this is an obsolete form factor/chip combination. I don't want a refurb or a system pull because, quite honestly, once I deal with this mess, I don't want to go through it again in another year or two. Find an Intel uBTX mobo and a (relatively -- hah, I still want at least a dual-core) inexpensive Intel CPU. At this point, the only things stock in the machine would be the case and the PSU. :) Buy a bare-bones kit (mobo/proc/PSU/case, sometimes even RAM) from somewhere like CompUSA/TigerDirect or Fry's and move all of the other hardware over. This makes life difficult because the copy of Vista is an upgrade, tied to the copy of XP which shipped on the Gateway, which is OEM and won't install on the new box. :) If I change the CPU brand (AMD to Intel), will I need to reinstall Windows, or can it just be reactivated? Where can I actually find a new, in-box, not system pull, not refurb AM2 uBTX mobo? Do they even exist anymore? What kind of money are we talking (US dollars)? The end goal is to get the machine functional again as cheaply as humanly possible. If it were my own machine, I wouldn't even be asking this, I'd be custom-building a new one. However, it's not mine, I'm shelling out of pocket for the fix (plus the work), and thus want to keep that end price low-low-low.

    Read the article

  • swapping or thrashing with vast amounts of unmapped pagecache

    - by Marco
    EDIT: I noticed that this is more appropriate for superuser.com, I apologize. I don't know how to delete this question. I'm using kubuntu jaunty (i386 32bit), kernel 2.6.28-13-generic. I've 4Gb of RAM, of which only 3317Mb are seen by the system (I guess because of the 32bit system). I'm seeing that the pagecache utilization is continually growing, up to the point that the system is unusable (after a few days). This happens also when I don't do anything (all user applications closed and the bare minimum of services enabled). If enabled, the system starts to use swap space (using it all in the end). Even if swap is disabled, disk activity becomes continuous, with the system unresponsive. For example, right now the system is working (albeit a tad slow), with only firefox and wing ide running, and I have 2Gb cached with only 45Mb mapped: $ free total used free shared buffers cached Mem: 3346388 3247328 99060 0 8416 2117980 -/+ buffers/cache: 1120932 2225456 Swap: 2144668 519448 1625220 $ cat /proc/meminfo MemTotal: 3346388 kB MemFree: 97128 kB Buffers: 7872 kB Cached: 2120224 kB SwapCached: 413860 kB Active: 2304596 kB Inactive: 865984 kB Active(anon): 2279168 kB Inactive(anon): 830236 kB Active(file): 25428 kB Inactive(file): 35748 kB Unevictable: 32 kB Mlocked: 32 kB HighTotal: 2492940 kB HighFree: 5456 kB LowTotal: 853448 kB LowFree: 91672 kB SwapTotal: 2144668 kB SwapFree: 1625244 kB Dirty: 84 kB Writeback: 0 kB AnonPages: 629304 kB Mapped: 45768 kB Slab: 45600 kB SReclaimable: 21756 kB SUnreclaim: 23844 kB PageTables: 4468 kB NFS_Unstable: 0 kB Bounce: 0 kB WritebackTmp: 0 kB CommitLimit: 3817860 kB Committed_AS: 3735020 kB VmallocTotal: 122880 kB VmallocUsed: 9352 kB VmallocChunk: 66600 kB HugePages_Total: 0 HugePages_Free: 0 HugePages_Rsvd: 0 HugePages_Surp: 0 Hugepagesize: 4096 kB DirectMap4k: 16376 kB DirectMap4M: 888832 kB If I try to drop the caches, little happes: # sync ; echo 3 > /proc/sys/vm/drop_caches ; free total used free shared buffers cached Mem: 3346388 3220580 125808 0 3020 2100600 -/+ buffers/cache: 1116960 2229428 Swap: 2144668 519356 1625312 Right now I've vm.swappiness = 5, but I've tried also with 0 and 1 (without noticeable differences). I've also tried vm.vfs_cache_pressure = 50 and 150 (again, no differences). As I said the pagecache eats all memory even with swapping turned off. What is happening? How to avoid this? TIA, Marco

    Read the article

  • SSD Fresh Does Not Start

    - by Jim Fell
    I recently installed a new 60GB SSD as my primary hard drive and re-installed Windows 7 Professional 64-bit. I then installed SSD Fress from Abelssoft to optimize Windows to run on the SSD. It seemed to install okay, but when I try to run the utility, its splash screen appears briefly before it quietly closes. No errors are displayed; the utility just fails to launch. I have run SSD Fresh on another SSD-equipped Windows 7 Pro x64 computer in the past without any problems. Does anyone know what might be preventing the program from running? I tried shutting down the Spybot Resident and disabling the firewall and virus scanner with no luck. I also tried running the tool as administrator; I even tried reinstalling it, running the installer as administrator. No luck. Every time I try to launch the program the Event Viewer logs this same set of errors: Error 4/2/2012 11:35:44 PM Application Error 1000 (100) Error 4/2/2012 11:35:43 PM .NET Runtime 1026 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None Error 4/2/2012 11:35:39 PM SideBySide 59 None For those who are interested, here is my system configuration: ASRock M3A770DE AM3 AMD 770 ATX AMD Motherboard AMD Athlon II X3 455 Rana 3.3GHz Socket AM3 95W Triple-Core Desktop Processor ADX455WFGMBOX G.SKILL Value Series 8GB (2 x 4GB) 240-Pin DDR3 SDRAM DDR3 1333 (PC3 10600) Desktop Memory Model F3-10600CL9D-8GBNT Mushkin Enhanced Chronos Deluxe MKNSSDCR60GB-DX 2.5" 60GB SATA III Synchronous MLC Internal Solid State Drive (SSD) (Primary/Boot HD) Western Digital Caviar Blue RFHWD1600AAJS 160GB 7200 RPM SATA 3.0Gb/s 3.5" Internal Hard Drive -Bare Drive (Secondary HD) Sony Optiarc CD/DVD Burner Black SATA Model AD-7261S-0B LightScribe Support RAIDMAX RX-850AE 850W ATX12V v2.3 / EPS12V SLI Certified CrossFire Ready 80 PLUS GOLD Certified Modular Active PFC Power Supply ASUS HD7850-DC2-2GD5 Radeon HD 7850 2GB 256-bit GDDR5 PCI Express 3.0 x16 HDCP Ready CrossFireX Support Video Card Asus ML228H 21.5" Full HD LED BackLight LED Monitor Slim Design (x3)

    Read the article

  • Crashes and freezes after fixing "BOOTMGR is missing" error

    - by Greg-J
    I came back from a 3-day weekend to a computer that was off. I leave my PC on 24/7, so this was odd. Turn it on to get the dreaded "BOOTMGR is missing" screen. Two attempts at Windows Recovery and it booted into Windows fine. After an hour or so, I get a frozen Chrome and my start bar disappears. Ctrl+Alt+Del brings up an error box telling me that Ctrl+Alt+Del failed to work properly. Clicking on any open application triggers an error (I can't recall the error now, but it essentially just said that the application couldn't be found running or something along those lines). I restart, and again, the same thing happens after a while of use. I turn it on, install the 47 updates I have or so, and then restart it. After a while of use (under an hour), it just freezes completely. My thoughts are: SSDs, RAM or PS. My system specs below: (RAID0) 2 x Crucial M4 CT128M4SSD2 2.5" 128GB SATA III MLC Internal Solid State Drive (SSD) CORSAIR Vengeance 16GB (4 x 4GB) 240-Pin DDR3 SDRAM DDR3 1600 (PC3 12800) Desktop Memory Model CML16GX3M4A1600C9 CORSAIR HX Series HX750 750W ATX12V 2.3 / EPS12V 2.91 SLI Ready CrossFire Ready 80 PLUS GOLD Certified Modular Active 1 x ASUS Maximus IV Gene-Z/GEN3 LGA 1155 Intel Z68 HDMI SATA 6Gb/s USB 3.0 Micro ATX Intel Motherboard 1 x Hitachi GST Deskstar 7K1000.C 0F10383 1TB 7200 RPM SATA 3.0Gb/s 3.5" Internal Hard Drive -Bare Drive 1 x Intel Core i7-2600K Sandy Bridge 3.4GHz (3.8GHz Turbo Boost) LGA 1155 95W Quad-Core Desktop Processor Intel HD Graphics 1 x SAPPHIRE 21197-00-40G Radeon HD 7970 3GB 384-bit GDDR5 PCI Express 3.0 x16 HDCP Ready CrossFireX Support Video Card 1 x Noctua NH-D14 120mm & 140mm SSO CPU Cooler This is all crammed in a pretty small case (NZXT Vulcan) and has been running perfectly problem-free since January. The only thing out of the ordinary is that there is a fan in the case that is now making noise whereas the case has previously been completely silent. I have no reason to believe this is anything more then correlation, but felt it is worth mentioning. I believe it MAY be the SSDs simply because of the BOOTMGR error, but not sure how to test that theory. My belief that it may be the RAM is simply from experience with frozen machines. I haven't had the time to memtest it, but will. The PS being the culprit is something I've picked up by reading similar threads on various forums, and it seems plausible. I am unsure how to test this though. ANY insight whatsover would be greatly appreciated!

    Read the article

  • Unecrypted Image of Truecrypt-Encrypted System Partition

    - by Dexter
    The general tenor around the internet seems to be that you can't create images of system partitions that have been encrypted (with truecrypt) other than with dd or similar sector-by-sector copy tools. These files however are very impractical given their size (and are obviously incompressible) which makes keeping multiple states/backups of your system partition rather expensive (..especially considering current hdd prices). The problem is that backup tools (like Acronis True Image, Clonezilla, etc.) won't give you the option to create an image of (mounted/opened) Truecrypt partitions, or that there is no recovery environment for restoring the backup, that would allow to run truecrypt before doing any actual restoring. After some trial and error however, I believe I have found a very simple way. Since Truecrypt (running in Linux) creates a virtual block device, that it uses for mounting the unencrypted partitions into the file system, partclone can be used for creating/restoring images. What I did: boot up a linux live disk mount/open the drive/device/partition in truecrypt unmount the filesystem mount point again, like so: umount /media/truecryptX ("X" being the partition number assigend by truecrypt) use partclone (this is what clonezilla would do too, except that clonezilla only offers you to back up real drive partitions, not virtual block devices): partclone.ntfs -c -s /dev/mapper/truecryptX -o nameOfBackupFile for restoring steps 1-3 remain the same, and step 4 is partclone.ntfs -r -s nameOfBackupFile -o /dev/mapper/truecryptX A backup and test-restore of the system (with this method) seems to have worked fine (and the changed settings were reverted to the backup-state). The backup file is ~40 GB (and compressible down to <8GB with 7zip/LZMA2 on the "fast" setting). I can't quite believe that I'm the only one that wants to create images of encrypted drives, but doesn't want to waste 100GB on the backup of one single system state. So my question now is, given how simple this was, and that no one seems to mention anywhere that this is possible - did I miss something? or did I do something wrong? Is there any situation that I didn't think of where this method will fail? Obviously, the backup file needs to be stored in some other encrypted place in order to still remain confidential, since it is unencrypted. Also, in order to do a full "bare metal" restore, one would have to actually first (re-)install Windows, encrypt it, and only then restore the backup file. The funny thing however is that you won't need to backup any partition tables, etc. since the reinstall will effectively take care of that. Is there anything else? This is imho still a lot better than having sector-by-sector images..

    Read the article

  • Sharing folder in a Virtual Private Windows Server 2008 R2 ?

    - by Triztian
    See Edit 2: Hello all, seems my involvement with computers has grown and I've found my self in the need to access a shared folder on a server. I've read some documentation and managed to set up the folder as a share, for this I created a local group and for now just one local user that has access to the share, the folder is in the public user folder and it's permissions should be (and I believe they are) read/write. The problem is that I can't connect from a remote machine I mean I don't know how the way it should be accessed, the server has a public IP and we use it also as a host to our website I don't know if that affects it though, the folder will be used as the "keeper" for the QuickBooks company files and has the database server manager installed. I've tried setting up a VPN Connection to the but no success. The server has a domain name a "http://www.example.com" that redirects to our website, I am unsure if it could be accessed that way, also the share has a location displayed when I right-click properties Heres what I've tried Setting up a VPN Connection (Windows Vista and 7) Got to the point where I got asked for credential and entered the user I created (which is not an admin) but I got a "Connection fail error 800" I suppose this is because in the domain field I entered the servers workgroup. right-click add network connection (Windows 7) Went through the wizard until I reached the point of entering the location, tried many things, the name in the share's properties(\\SOMETHING\Share), the http://www.example.com , the IP address I'm quite unfamiliar with this, so I have my guesses: Since the group and user are local they do not have access to the folder. The firewall in the server is blocking my connection. Anyways, any help and guidence is truly appreciated. EDIT 1: As @tony roth pointed out it may be a security fail, an I commented it out to management and said that that is not an issue, so please bare with me. EDIT 2: I've found out that the real question could be streamlined to "Sharing folder in a Virtual Private Server?", as thats what we have, a virtual private windows server 2008 R2, and I would like to know how to make it show like a normal folder in the client computer. Thanks again for all of your support.

    Read the article

  • Recognizing Dell EquilLogic with Nagios

    - by user3677595
    EDIT: All firmware and models are compatible, that is why nothing is posted about it. Okay, so there will be a lot here, so please bare with me. I've been working on this now for a few hours (reading manuals and such) so I'm not just coming here right out of the blue. I am working on a PRE-EXISTING Nagios server where there are several other existing plugins and checks running and working. Now I want to add another server there to check so I made the following modifications: First and foremost, I added a file to /usr/local/nagios/libexec named: check_equallogic.sh. The permissions are 755, the same as all others. I have chowned to nagios:nagios and in the listing it shows the Owner as Nagios. I then added a command to the commands.cfg file in \usr\local\nagios\etc\objects that shows the following: # 'check_equallogic' command definition define command{ command_name check_equallogic command_line $USER1$/check_equallogic -H $HOSTADDRESS$ -C $ARG1$ -t $ARG2$ $ARG3$ } Following this, I created a file named equallogic.cfg in the objects directory and it contains (more or less): define host{ use linux-server ; Inherit default values from a template host_name 172.16.50.11 ; The name we're giving to this device alias EqualLogic ; A longer name associated with the device address 172.16.50.11 ; IP address of the device contact_groups admins } Check Equallogic Information define service{ use generic-service host_name 172.16.50.11 service_description General Information check_command check_equallogic!public!info } After ensuring that permissions are okay for all files, I restart the nagios service, no errors. When I go into the WebGUI, I get the following errors AFTER the check runs: (Return code of 127 is out of bounds - plugin may be missing) Extra, probably unrelated problem Furthermore, when I log into the EquilLogic server, under Audit logs I get the following error: Level: AUDIT Time: 26/05/2014 3:59:13 PM Member: ps4100-1 Subsystem: agent Event ID: 22.7.1 SNMP packet validation failed, request received from 172.16.10.11 An snmpwalk receives a timeout, whereas others succeed. I will work on importing the MIBs tomorrow. The reason why I am mentioning it is because I want to make sure that it is only a MIB issue for the SNMP. If it is, then ignore this area. I am entirely unsure of what to do here.

    Read the article

  • .htaccess ignored, SPECIFIC to EC2 - not the usual suspects

    - by tedneigerux
    I run 8-10 EC2 based web servers, so my experience is many hours, but is limited to CentOS; specifically Amazon's distribution. I'm installing Apache using yum, so therefore getting Amazon's default compilation of Apache. I want to implement canonical redirects from non-www (bare/root) domain to www.domain.com for SEO using mod_rewrite BUT MY .htaccess FILE IS CONSISTENTLY IGNORED. My troubleshooting steps (outlined below) lead me to believe it's something specific to Amazon's build of Apache. TEST CASE Launch a EC2 Instance, e.g. Amazon Linux AMI 2013.03.1 SSH to the Server Run the commands: $ sudo yum install httpd $ sudo apachectl start $ sudo vi /etc/httpd/conf/httpd.conf $ sudo apachectl restart $ sudo vi /var/www/html/.htaccess In httpd.conf I changed the following, in the DOCROOT section / scope: AllowOverride All In .htaccess, added: (EDIT, I added RewriteEngine On later) RewriteCond %{HTTP_HOST} ^domain\.com$ [NC] RewriteRule ^/(.*) http://www.domain.com/$1 [R=301,L] Permissions on .htaccess are correct, AFAI can tell: $ ls -al /var/www/html/.htaccess -rwxrwxr-x 1 git apache 142 Jun 18 22:58 /var/www/html/.htaccess Other info: $ httpd -v Server version: Apache/2.2.24 (Unix) Server built: May 20 2013 21:12:45 $ httpd -M Loaded Modules: core_module (static) ... rewrite_module (shared) ... version_module (shared) Syntax OK EXPECTED BEHAVIOR $ curl -I domain.com HTTP/1.1 301 Moved Permanently Date: Wed, 19 Jun 2013 12:36:22 GMT Server: Apache/2.2.24 (Amazon) Location: http://www.domain.com/ Connection: close Content-Type: text/html; charset=UTF-8 ACTUAL BEHAVIOR $ curl -I domain.com HTTP/1.1 200 OK Date: Wed, 19 Jun 2013 12:34:10 GMT Server: Apache/2.2.24 (Amazon) Connection: close Content-Type: text/html; charset=UTF-8 TROUBLESHOOTING STEPS In .htaccess, added: BLAH BLAH BLAH ERROR RewriteCond %{HTTP_HOST} ^domain\.com$ [NC] RewriteRule ^/(.*) http://www.domain.com/$1 [R=301,L] My server threw an error 500, so I knew the .htaccess file was processed. As expected, it created an Error log entry: [Wed Jun 19 02:24:19 2013] [alert] [client XXX.XXX.XXX.XXX] /var/www/html/.htaccess: Invalid command 'BLAH BLAH BLAH ERROR', perhaps misspelled or defined by a module not included in the server configuration Since I have root access on the server, I then tried moving my rewrite rule directly to the httpd.conf file. THIS WORKED. This tells us several important things are working. $ curl -I domain.com HTTP/1.1 301 Moved Permanently Date: Wed, 19 Jun 2013 12:36:22 GMT Server: Apache/2.2.24 (Amazon) Location: http://www.domain.com/ Connection: close Content-Type: text/html; charset=UTF-8 HOWEVER, it is bothering me that it didn't work in the .htaccess file. And I have other use cases where I need it to work in .htaccess (e.g. an EC2 instance with named virtual hosts). Thank you in advance for your help.

    Read the article

  • How should we serve files in a small bioinformatics cluster?

    - by cespinoza
    We have a small cluster of six ubuntu servers. We run bioinformatics analyses on these clusters. Each analysis takes about 24 hours to complete, each core i7 server can handle 2 at a time, takes as input about 5GB data and outputs about 10-25GB of data. We run dozens of these a week. The software is a hodgepodge of custom perl scripts and 3rd party sequence alignment software written in C/C++. Currently, files are served from two of the compute nodes (yes, we're using compute nodes as file servers)-- each node has 5 1TB sata drives mounted separately (no raid) and is pooled via glusterfs 2.0.1. They each have as 3 bonded intel ethernet pci gigabit ethernet cards, attached to a d-link DGS-1224T switch ($300 24 port consumer-level). We are not currently using jumbo frames (not sure why, actually). The two file-serving compute nodes are then mirrored via glusterfs. Each of the four other nodes mounts the files via glusterfs. The files are all large (4gb+), and are stored as bare files (no database/etc) if that matters. As you can imagine, this is a bit of a mess that grew organically without forethought and we want to improve it now that we're running out of space. Our analyses are I/O intensive and it is a bottle neck-- we're only getting 140mB/sec between the two fileservers, maybe 50mb/sec from the clients (which only have single NICs). We have a flexible budget which I can probably get up $5k or so. How should we spend our budget? We need at least 10TB of storage fast enough to serve all nodes. How fast/big does the cpu/memory of such a file server have to be? Should we use NFS, ATA over Ethernet, iSCSI, Glusterfs, or something else? Should we buy two or more servers and create some sort of storage cluster, or is 1 server enough for such a small number of nodes? Should we invest in faster NICs (say, PCI-express cards with multiple connectors)? The switch? Should we use raid, if so, hardware or software? and which raid (5, 6, 10, etc)? Any ideas appreciated. We're biologists, not IT gurus.

    Read the article

  • “Being Agile” Means No Documentation, Right?

    - by jesschadwick
    Ask most software professionals what Agile is and they’ll probably start talking about flexibility and delivering what the customer wants.  Some may even mention the word “iterations”.  But inevitably, they’ll say at some point that it means less or even no documentation.  After all, doesn’t creating, updating, and circulating painstakingly comprehensive documentation that everyone and their mother have officially signed off on go against the very core of Agile?  Of course it does!  But really, they’re missing the point! Read The Agile Manifesto. (No, seriously - read it now. It’s short. I’ll wait.)  It’s essentially a list of values.  More specifically, it’s a right-side/left-side weighted list of values:  “Value this over that”. Many people seem to get the impression that this is really a “good vs. bad” list and that those values on the right side are evil and should essentially be tossed on the floor.  This leads to the conclusion that in order to be Agile we must throw away our fancy expensive tools, document as little as possible, and scoff at the idea of a project plan.  This conclusion is quite convenient because it essentially means “less work, more productivity!” (particularly in regards to the documentation and project planning).  I couldn’t disagree with this conclusion more. My interpretation of the Manifesto targets “over” as the operative word.  It’s not just a list of right vs. wrong or good vs. bad.  It’s a list of priorities.  In other words, none of the concepts on the list should be removed from your development lifecycle – they are all important… just not equally important.  This is not a unique interpretation, in fact it says so right at the end of the manifesto! So, the next time your team sits down to tackle that big new project, don’t make the first order of business to outlaw all meetings, documentation, and project plans.  Instead, collaborate with both your team and the business members involved (you do have business members sitting in the room, directly involved in the project planning, right?) and determine the bare minimum that will allow all of you to work and communicate in the best way possible.  This often means that you can pick and choose which parts of the Agile methodologies and process work for your particular project and end up with an amalgamation of Waterfall, Agile, XP, SCRUM and whatever other methodologies the members of your team have been exposed to (my favorite is “SCRUMerfall”). The biggest implication of this is that there is no one way to implement Agile.  There is no checklist with which you can tick off boxes and confidently conclude that, “Yep, we’re Agile™!”  In fact, depending on your business and the members of your team, moving to Agile full-bore may actually be ill-advised.  Such a drastic change just ends up taking everyone out of their comfort zone which they inevitably fall back into by the end of the project.  This often results in frustration to the point that Agile is abandoned altogether because “we just need to ship something!”  Needless to say, this is far more devastating to a project. Instead, I offer this approach: keep it simple and take it slow.  If your business members or customers are only involved at the beginning phases and nowhere to be seen until the project is delivered, invite them to your daily meetings; encourage them to keep up to speed on what’s going on on a daily basis and provide feedback.  If your current process is heavy on the documentation, try to reduce it as opposed to eliminating it outright.  If you need a “TPS Change Request” signed in triplicate with a 5-day “cooling off period” before a change is implemented, try a simple bug tracking system!  Tighten the feedback loop! Finally, at the end of every “iteration” (whatever that means to you, as long as it’s relatively frequent), take as much time as you can spare (even if it’s an hour or so) and perform some kind of retrospective.  Learn from your mistakes.  Figure out what’s working for you and what’s not, then fix it.  Before you know it you’ve got a handful of iterations and/or projects under your belt and you sit down with your team to realize that, “Hey, this is working - we’re pretty Agile!”  After all, Agile is a Zen journey.  It’s a destination that you aim for, not force, and even if you never reach true “enlightenment” that doesn’t mean your team can’t be exponentially better off from merely taking the journey.

    Read the article

  • PASS Summit 2011 &ndash; Part III

    - by Tara Kizer
    Well we’re about a month past PASS Summit 2011, and yet I haven’t finished blogging my notes! Between work and home life, I haven’t been able to come up for air in a bit.  Now on to my notes… On Thursday of the PASS Summit 2011, I attended Klaus Aschenbrenner’s (blog|twitter) “Advanced SQL Server 2008 Troubleshooting”, Joe Webb’s (blog|twitter) “SQL Server Locking & Blocking Made Simple”, Kalen Delaney’s (blog|twitter) “What Happened? Exploring the Plan Cache”, and Paul Randal’s (blog|twitter) “More DBA Mythbusters”.  I think my head grew two times in size from the Thursday sessions.  Just WOW! I took a ton of notes in Klaus' session.  He took a deep dive into how to troubleshoot performance problems.  Here is how he goes about solving a performance problem: Start by checking the wait stats DMV System health Memory issues I/O issues I normally start with blocking and then hit the wait stats.  Here’s the wait stat query (Paul Randal’s) that I use when working on a performance problem.  He highlighted a few waits to be aware of such as WRITELOG (indicates IO subsystem problem), SOS_SCHEDULER_YIELD (indicates CPU problem), and PAGEIOLATCH_XX (indicates an IO subsystem problem or a buffer pool problem).  Regarding memory issues, Klaus recommended that as a bare minimum, one should set the “max server memory (MB)” in sp_configure to 2GB or 10% reserved for the OS (whichever comes first).  This is just a starting point though! Regarding I/O issues, Klaus talked about disk partition alignment, which can improve SQL I/O performance by up to 100%.  You should use 64kb for NTFS cluster, and it’s automatic in Windows 2008 R2. Joe’s locking and blocking presentation was a good session to really clear up the fog in my mind about locking.  One takeaway that I had no idea could be done was that you can set a timeout in T-SQL code view LOCK_TIMEOUT.  If you do this via the application, you should trap error 1222. Kalen’s session went into execution plans.  The minimum size of a plan is 24k.  This adds up fast especially if you have a lot of plans that don’t get reused much.  You can use sys.dm_exec_cached_plans to check how often a plan is being reused by checking the usecounts column.  She said that we can use DBCC FLUSHPROCINDB to clear out the stored procedure cache for a specific database.  I didn’t know we had this available, so this was great to hear.  This will be less intrusive when an emergency comes up where I’ve needed to run DBCC FREEPROCCACHE. Kalen said one should enable “optimize for ad hoc workloads” if you have an adhoc loc.  This stores only a 300-byte stub of the first plan, and if it gets run again, it’ll store the whole thing.  This helps with plan cache bloat.  I have a lot of systems that use prepared statements, and Kalen says we simulate those calls by using sp_executesql.  Cool! Paul did a series of posts last year to debunk various myths and misconceptions around SQL Server.  He continues to debunk things via “DBA Mythbusters”.  You can get a PDF of a bunch of these here.  One of the myths he went over is the number of tempdb data files that you should have.  Back in 2000, the recommendation was to have as many tempdb data files as there are CPU cores on your server.  This no longer holds true due to the numerous cores we have on our servers.  Paul says you should start out with 1/4 to 1/2 the number of cores and work your way up from there.  BUT!  Paul likes what Bob Ward (twitter) says on this topic: 8 or less cores –> set number of files equal to the number of cores Greater than 8 cores –> start with 8 files and increase in blocks of 4 One common myth out there is to set your MAXDOP to 1 for an OLTP workload with high CXPACKET waits.  Instead of that, dig deeper first.  Look for missing indexes, out-of-date statistics, increase the “cost threshold for parallelism” setting, and perhaps set MAXDOP at the query level.  Paul stressed that you should not plan a backup strategy but instead plan a restore strategy.  What are your recoverability requirements?  Once you know that, now plan out your backups. As Paul always does, he talked about DBCC CHECKDB.  He said how fabulous it is.  I didn’t want to interrupt the presentation, so after his session had ended, I asked Paul about the need to run DBCC CHECKDB on your mirror systems.  You could have data corruption occur at the mirror and not at the principal server.  If you aren’t checking for data corruption on your mirror systems, you could be failing over to a corrupt database in the case of a disaster or even a planned failover.  You can’t run DBCC CHECKDB against the mirrored database, but you can run it against a snapshot off the mirrored database.

    Read the article

  • Imaging: Paper Paper Everywhere, but None Should be in Sight

    - by Kellsey Ruppel
    Author: Vikrant Korde, Technical Architect, Aurionpro's Oracle Implementation Services team My wedding photos are stored in several empty shoeboxes. Yes...I got married before digital photography was mainstream...which means I'm old. But my parents are really old. They have shoeboxes filled with vacation photos on slides (I doubt many of you have even seen a home slide projector...and I hope you never do!). Neither me nor my parents should have shoeboxes filled with any form of photographs whatsoever. They should obviously live in the digital world...with no physical versions in sight (other than a few framed on our walls). Businesses grapple with similar challenges. But instead of shoeboxes, they have file cabinets and warehouses jam packed with paper invoices, legal documents, human resource files, material safety data sheets, incident reports, and the list goes on and on. In fact, regulatory and compliance rules govern many industries, requiring that this paperwork is available for any number of years. It's a real challenge...especially trying to find archived documents quickly and many times with no backup. Which brings us to a set of technologies called Image Process Management (or simply Imaging or Image Processing) that are transforming these antiquated, paper-based processes. Oracle's WebCenter Content Imaging solution is a combination of their WebCenter suite, which offers a robust set of content and document management features, and their Business Process Management (BPM) suite, which helps to automate business processes through the definition of workflows and business rules. Overall, the solution provides an enterprise-class platform for end-to-end management of document images within transactional business processes. It's a solution that provides all of the capabilities needed - from document capture and recognition, to imaging and workflow - to effectively transform your ‘shoeboxes’ of files into digitally managed assets that comply with strict industry regulations. The terminology can be quite overwhelming if you're new to the space, so we've provided a summary of the primary components of the solution below, along with a short description of the two paths that can be executed to load images of scanned documents into Oracle's WebCenter suite. WebCenter Imaging (WCI): the electronic document repository that provides security, annotations, and search capabilities, and is the primary user interface for managing work items in the imaging solution SOA & BPM Suites (workflow): provide business process management capabilities, including human tasks, workflow management, service integration, and all other standard SOA features. It's interesting to note that there a number of 'jumpstart' processes available to help accelerate the integration of business applications, such as the accounts payable invoice processing solution for E-Business Suite that facilitates the processing of large volumes of invoices WebCenter Enterprise Capture (WEC): expedites the capture process of paper documents to digital images, offering high volume scanning and importing from email, and allows for flexible indexing options WebCenter Forms Recognition (WFR): automatically recognizes, categorizes, and extracts information from paper documents with greatly reduced human intervention WebCenter Content: the backend content server that provides versioning, security, and content storage There are two paths that can be executed to send data from WebCenter Capture to WebCenter Imaging, both of which are described below: 1. Direct Flow - This is the simplest and quickest way to push an image scanned from WebCenter Enterprise Capture (WEC) to WebCenter Imaging (WCI), using the bare minimum metadata. The WEC activities are defined below: The paper document is scanned (or imported from email). The scanned image is indexed using a predefined indexing profile. The image is committed directly into the process flow 2. WFR (WebCenter Forms Recognition) Flow - This is the more complex process, during which data is extracted from the image using a series of operations including Optical Character Recognition (OCR), Classification, Extraction, and Export. This process creates three files (Tiff, XML, and TXT), which are fed to the WCI Input Agent (the high speed import/filing module). The WCI Input Agent directory is a standard ingestion method for adding content to WebCenter Imaging, the process for doing so is described below: WEC commits the batch using the respective commit profile. A TIFF file is created, passing data through the file name by including values separated by "_" (underscores). WFR completes OCR, classification, extraction, export, and pulls the data from the image. In addition to the TIFF file, which contains the document image, an XML file containing the extracted data, and a TXT file containing the metadata that will be filled in WCI, are also created. All three files are exported to WCI's Input agent directory. Based on previously defined "input masks", the WCI Input Agent will pick up the seeding file (often the TXT file). Finally, the TIFF file is pushed in UCM and a unique web-viewable URL is created. Based on the mapping data read from the TXT file, a new record is created in the WCI application.  Although these processes may seem complex, each Oracle component works seamlessly together to achieve a high performing and scalable platform. The solution has been field tested at some of the largest enterprises in the world and has transformed millions and millions of paper-based documents to more easily manageable digital assets. For more information on how an Imaging solution can help your business, please contact [email protected] (for U.S. West inquiries) or [email protected] (for U.S. East inquiries). About the Author: Vikrant is a Technical Architect in Aurionpro's Oracle Implementation Services team, where he delivers WebCenter-based Content and Imaging solutions to Fortune 1000 clients. With more than twelve years of experience designing, developing, and implementing Java-based software solutions, Vikrant was one of the founding members of Aurionpro's WebCenter-based offshore delivery team. He can be reached at [email protected].

    Read the article

  • Community Branching

    - by Dane Morgridge
    As some may have noticed, I have taken a liking to Ruby (and Rails in particular) quite a bit recently. This last weekend I spoke at the NYC Code Camp on a comparison of ASP.NET and Rails as well as an intro to Entity Framework talk.  I am speaking at RubyNation in April and have submitted to other ruby conferences around the area and I am also doing a Rails and MongoDB talk at the Philly Code Camp in April. Before you start to think this is my "I'm leaving .NET post", which it isn't so I need to clarify. I am not, nor do I intend to any time in the near future plan on abandoning .NET.  I am simply branching out into another community based on a development technology that I very much enjoy.  If you look at my twitter bio, you will see that I am into Entity Framework, Ruby on Rails, C++ and ASP.NET MVC, and not necessarily in that order.  I know you're probably thinking to your self that I am crazy, which is probably true on several levels (especially the C++ part). I was actually crazy enough at the NYC Code Camp to show up wearing a Linux t-shirt, presenting with my MacBook Pro on Entity Framework, ASP.NET MVC and Rails. (I did get pelted in the head with candy by Rachel Appel for it though) At all of the code camps I am submitting to this year, i will be submitting sessions on likely all four topics, and some sessions will be a combination of 2 or more.  For example, my "ASP.NET MVC: A Gateway To Rails?" talk touches ASP.NET MVC, Entity Framework Code First and Rails. Simply put (and I talk about this in my MVC & Rails talk) is that learning and using Rails has made me a better ASP.NET MVC developer. Just one example of this is helper methods.  When I started working with ASP.NET MVC, I didn't really want to use helpers and preferred to just use standard html tags, especially where links were concerned.  It was just me being stubborn and not really seeing all of the benefit of the helpers.  To my defense, coming from WebForms, I wanted to be as bare metal as possible and it seemed at first like a lot of the helpers were an unnecessary abstraction. I took my first look at Rails back in v1 and didn't spend very much time with it so I dismissed it and went on my merry ASP.NET WebForms way.  Then I picked up ASP.NET MVC and grasped the MVC pattern itself much better. After this, I took another look at Rails and everything made sense.  I decided then to learn Rails. (I think it is important for developers to learn new languages and platforms regularly so it was a natural progression for me) I wanted to learn it the right way, so when I dug into code, everyone used helpers everywhere for pretty much everything possible. I took some time to dig in and found out how helpful they were and subsequently realized how awesome they were in ASP.NET MVC also and started using them. In short, I love Rails (and Ruby in general).  I also love ASP.NET MVC and Entity Framework and yes I still love C++.  I have varying degrees of love for them individually at any given moment and it is likely to shift based on the current project I am working on.  I know you're thinking it so before you ask the question. "Which do I use when?", I'm going to give the standard developer answer of: It depends.  There are a lot of factors that I am not going to even go into that would go into a decision.  The most basic question I would ask though is,  does this project depend on .NET?  If it does, then I'd say that ASP.NET MVC is probably going to be the more logical choice and I am going to leave it at that.  I am working on projects right now in both technologies and I don't see that changing anytime soon (one project even uses both). With all that being said, you'll find me at code camps, conferences and user groups presenting on .NET, Ruby or both, writing about .NET and Ruby and I will likely be blogging on both in the future.  I know of others that have successfully branched out to other communities and with any luck I'll be successful at it too. On a (sorta) side note, I read a post by Justin Etheredge the other day that pretty much sums up my feelings about Ruby as a language.  I highly recommend checking it out: What Is So Great About Ruby?

    Read the article

  • My Red Gate Experience

    - by Colin Rothwell
    I’m Colin, and I’ve been an intern working with Mike in publishing on Simple-Talk and SQLServerCentral for the past ten weeks. I’ve mostly been working “behind the scenes”, making improvements to the spam filtering, along with various other small tweaks. When I arrived at Red Gate, one of the first things Mike asked me was what I wanted to get out of the internship. It wasn’t a question I’d given a great deal of thought to, but my immediate response was the same as almost anybody: to support my growing family. Well, ok, not quite that, but money was certainly a motivator, along with simply making sure that I didn’t get bored over the summer. Three months is a long time to fill, and many of my friends end up getting bored, or worse, knitting obsessively. With the arrogance which seems fairly common among Cambridge people, I wasn’t expecting to really learn much here! In my mind, the part of the year where I am at Uni is the part where I learn things, whilst Red Gate would be an opportunity to apply what I’d learnt. Thankfully, the opposite is true: I’ve learnt a lot during my time here, and there has been a definite positive impact on the way I write code. The first thing I’ve really learnt is that test-driven development is, in general, a sensible way of working. Before coming, I didn’t really get it: how could you test something you hadn’t yet written? It didn’t make sense! My problem was seeing a test as having to test all the behaviour of a given function. Writing tests which test the bare minimum possible and building them up is a really good way of crystallising the direction the code needs to grow in, and ensures you never attempt to write too much code at time. One really good experience of this was early on in my internship when Mike and I were working on the query used to list active authors: I’d written something which I thought would do the trick, but by starting again using TDD we grew something which revealed that there were several subtle mistakes in the query I’d written. I’ve also been awakened to the value of pair programming. Whilst I could sort of see the point before coming, I also thought that it was impossible that two people would ever get more done at the same computer than if they were working separately. I still think that this is true for projects with pieces that developers can easily work on independently, and with developers who both know the codebase, but I’ve found that pair programming can be really good for learning a code base, and for building up small projects to the point where you can start working on separate components, as well as solving particularly difficult problems. Later on in my internship, for my down tools week project, I was working on adding Python support to Glimpse. Another intern and I we pair programmed the entire project, using ping pong pair programming as much as possible. One bonus that this brought which I wasn’t expecting was that I found myself less prone to distraction: with someone else peering over my shoulder, I didn’t have the ever-present temptation to open gmail, or facebook, or yammer, or twitter, or hacker news, or reddit, and so on, and so forth. I’m quite proud of this project: I think it’s some of the best code I’ve written. I’ve also been really won over to the value of descriptive variables names. In my pre-Red Gate life, as a lone-ranger style cowboy programmer, I’d developed a tendency towards laziness in variable names, sometimes abbreviating or, worse, using acronyms. I’ve swiftly realised that this is a bad idea when working with a team: saving a few key strokes is inevitably not worth it when it comes to reading code again in the future. Longer names also mean you can do away with a majority of comments. I appreciate that if you’ve come up with an O(n*log n) algorithm for something which seemed O(n^2), you probably want to explain how it works, but explaining what a variable name means is a big no no: it’s so very easy to change the behaviour of the code, whilst forgetting about the comments. Whilst at Red Gate, I took the opportunity to attend a code retreat, which really helped me to solidify all the things I’d learnt. To be completely free of any existing code base really lets you focus on best practises and think about how you write code. If you get a chance to go on a similar event, I’d highly recommend it! Cycling to Red Gate, I’ve also become much better at fitting inner tubes: if you’re struggling to get the tube out, or re-fit the tire, letting a bit of air out usually helps. I’ve also become quite a bit better at foosball and will miss having a foosball table! I’d like to finish off by saying thank you to everyone at Red Gate for having me. I’ve really enjoyed working with, and learning from, the team that brings you this web site. If you meet any of them, buy them a drink!

    Read the article

  • Career guidance/advice for Junior-level Software Engineer [closed]

    - by John Do
    I have quite a few questions on my mind, so please bare with me. Please don't feel obligated to answer all of them, any as you choose will do. I'd appreciate if you could share some insight on any of these. Before I begin, some context: I currently have almost two years of professional experience as a Software Engineer, mainly developing software in Java. At this point, I feel that I have reached the peak in my career growth at the current company I am at and therefore I am looking for a new job, ideally again, as a Software Engineer. I have been interviewing for the past few months casually but have not had luck with companies I have a passion for. So, in no particular order - 1) In general, what are your thoughts on having graduate degrees in CS / Software Engineering. How much does it influence a salary increase, and do you think it's beneficial when working on real-world problems? I get the sense that a graduate degree in the field is trivial unless you really have a passion for research. 2) In general, in professional practice, how often had you have to write your own data structures and "complex" algorithms from scratch? In my own work, I have found myself relying mainly on third-party frameworks and the Java standard library to implement solutions as per business requirements. What are your thoughts on this? 3) In terms of resume, I feel the most ambivalent here. I want to be able to "blemish" my resume to a certain extent so that it stands out from others', but at the same time I do not want to over-exagerate my abilities. How do you strike a balance here? For example: I say that I am proficient in Java with data structures and algorithms. This is obviously a subjective and relative statement. I've taken the classes in my undergrad, and I've applied it in my work experience. What I feel as "prociency" can be seen as junior-level to others. How do you know what to say? Most of the time, recruiters (with no technical background) will be looking for keywords that stand out. This leads me to my next question (4). 4) Just from interviewing for the past few months (and getting plenty of rejections), I've come to realize that I may not be as proficient in data structures and algorithms as I thought I was. Do you think it's a good idea to remove the "proficient in java/data structure and algorithms"? I feel that being too hoenst on the resume will impede me from scoring opportunities to even have an interview with top-notch companies. What are your thoughts? 5) What is the absolute "must-have" knowledge going into a technical interview? I've been practicing several algorithmic and data sturcture problems now, and I feel that my abilities to solve arbitrary problems efficiently has not gotten significantly better. Do you think these abilities are something innate - it's either you have in you, or you don't? How can you teach yourself to learn, if you will? 6) How easy is it to go from industry/function to the next? I work mainly with backend technologies and I'm now interested in working with the frontend, i.e javascript,jquery,php or even mobile development. In your own experience, how did you not get pidgeon holed in your career? I feel that the choices you make now ultimately decide your future. As cliche as it sounds, I think it may be true. Here's what I mean: you've worked mainly as a backend engineer, people are interested in you doing the same thing since you've already accumulated experience in that function. How do get experience in a new function if people won't accept you because you don't already have it? It's a catch 22, you see... Are side projects the only real way to help you move from one function to another that you're truly interested in? For example: I could start writing my own mobile applications, even though I've worked mainly on the backend. Thanks so much for the long read. As a relatively new engineer to the real world, I am very humble and would like those who are experienced to shed some light. Thank you so much.

    Read the article

  • Microsoft BUILD 2013 Day 1&ndash;Keynote

    - by Tim Murphy
    Originally posted on: http://geekswithblogs.net/tmurphy/archive/2013/06/27/microsoft-build-2013-day-1ndashkeynote.aspx This one is going to be a little long because the keynote was jam-packed so bare with me. The keynote for the first day of BUILD 2013 was kicked off by Steve Balmer.  He made it very clear that Microsoft’s focus is on accelerating its time to market with products and product updates.  His quote was that “Rapid release” is the new norm.  He continued by showing off several new Lumias that have been buzzing around the internet for a while and announce that Sprint will now be carrying the HTC 8XT and Samsung ATIV. Balmer is known for repeating words or phrase for affect.  This time it was “Rapid release, rapid release” and “Touch, touch, touch, touch, touch, …”.  This was fun, but even more fun was when he announce that all attendees would receive an Acer Iconia 8” tablet. SCORE! The next subject Balmer focused on is new apps.  The three new ones were Flipboard, Facebook and NFL Fantasy Football.  I liked the first two because these are ones that people coming from other platforms are missing.  The NFL app is great just because it targets a demographic that can be fanatical.  If these types of apps keep coming than the missing app argument goes away. While many Negative Nancy’s are describing Windows 8.1 as Windows 180 Steve Balmer chose to call it a “refined blend” as in a coffee that has been improved with a new mix.  This includes more multi-tasking options and leveraging Bing straight throughout the entire ecosystem. He ended this first section by explaining that this will also bring more Bing development opportunities to the community. Steve Balmer was followed by Julie Larson-Green who spent her time on stage selling us on Windows 8 all over again from my point of view.  Something that I would not have thought was needed until I had listened to some other attendees who had a number of concerns and complaints.  She showed a number of new gestures that will come with Windows 8.1, and while they were cool I was left wondering if they really improved the experience.  I guess only time will tell. I did like the fact that it the UI implementation to bring up “All Apps” now mirrors that of Windows Phone.  The consistency is a big step forward that I hope to see continue.  The cool factor went up from there as she swiped content from a desktop (mega-tablet) to the XBox One.  This seamless experience I believe is what is really needed for any future platform to be relevant. I was much more enthused by the presentation of Antoine Leblond who humbled us by letting us know that there are 5k new API.  How that can be or how anyone would ever use all of them is another question.  His announcement was that the Visual Studio 2013 preview would be available today along with the Windows 8.1 bits.  One of the features of VS2013 that he demonstrated is the power consumption profiler.  With battery life being a key factor with consumer consumption devices this is a welcome addition. He didn’t limit his presentation to VS2013 features though.  He showed how the Store has been redesigned to enable better search and discoverability of apps and how Win 8.1 can perform multiple screen scales depending on the resolution of the device automatically.  The last feature he demoed was the real time video streaming API which he made sure we understood by attaching a Surface to a little robot.  Oh, but there was one more thing.  Antoine and Julie announce that all attendees would also be getting Surface Pros.  BONUS! How much more could there be?  Gurdeep Singh Pall was about to pile on.  He introduced us to Bing as a platform (BaaP?).  He said if they (Microsoft) could do something with and API that is good 3rd party developers can do something that is dynamite and showed us some of the tools they had produced.  These included natural user interface improvements such as voice commands that looked to put Siri to shame.  Add to that 3D, OCR and translation capabilities and the future looks to be full of opportunities. Balmer then came out to show us one last thing.  Project Spark is a game design environment that will be available for Windows 8.1, XBox 360 and XBox One.  All I can say is that if my kids get their hands on this they are going to be able to learn some of what dad does in a much more enjoyable way. At the end of it all I was both exhausted and energized by what I saw.  What could they have possibly left for the day 2 keynote?  I hear it will feature Scott Hanselman.  If that is right we are in for a treat.  See you there. del.icio.us Tags: BUILD 2013,Windows 8.1,Winodws Phone,XAML,Keynote,Bing,Visual Studio 2013,Project Spark

    Read the article

  • The architecture and technologies to use for a secure, fast, reliable and easily scalable web application

    - by DSoul
    ^ For actual questions, skip to the lists down below I understand, that his is a vague topic, but please, before you turn the other way and disregard me, hear me out. I am currently doing research for a web application(I don't know if application is the correct word for it, but I will proceed w/ that for now), that one day might need to be everything mentioned in the title. I am bound by nothing. That means that every language, OS and framework is acceptable, but only if it proves it's usefulness. And if you are going to say, that scalability and speed depend on the code I write for this application, then I agree, but I am just trying to find something, that wouldn't stand in my way later on. I have done quite a bit reading on this subject, but I still don't have a clear picture, to what suits my needs, so I come to you, StackOverflow, to give me directions. I know you all must be wondering what I'm building, but I assure you, that it doesn't matter. I have heard of 12 factor app though, if you have any similar guidelines or what is, to suggest the please, go ahead. For the sake of keeping your answers as open as possible, I'm not gonna provide you my experience regarding anything written in this question. ^ Skippers, start here First off - the weights of the requirements are probably something like that (on a scale of 10): Security - 10 Speed - 5 Reliability (concurrency) - 7.5 Scalability - 10 Speed and concurrency are not a top priority, in the sense, that the program can be CPU intensive, and therefore slow, and only accept a not-that-high number of concurrent users, but both of these factors must be improvable by scaling the system Anyway, here are my questions: How many layers should the application have, so it would be future-proof and could best fulfill the aforementioned requirements? For now, what I have in mind is the most common version: Completely separated front end, that might be a web page or an MMI application or even both. Some middle-ware handling communication between the front and the back end. This is probably a server that communicates w/ the front end via HTTP. How the communication w/ the back end should be handled is probably dependent on the back end. The back end. Something that handles data through resources like DB and etc. and does various computations w/ the data. This, as the highest priority part of the software, must be easily spread to multiple computers later on and have no known security holes. I think ideally the middle-ware should send a request to a queue from where one of the back end processes takes this request, chops it up to smaller parts and buts these parts of the request back onto the same queue as the initial request, after what these parts will be then handled by other back end processes. Something *map-reduce*y, so to say. What frameworks, languages and etc. should these layers use? The technologies used here are not that important at this moment, you can ignore this part for now I've been pointed to node.js for this part. Do you guys know any better alternatives, or have any reasons why I should (not) use node.js for this particular job. I actually have no good idea, what to use for this job, there are too many options out there, so please direct me. This part (and the 2. one also, I think) depend a lot on the OS, so suggest any OSs alongside w/ the technologies/frameworks. Initially, all computers (or 1 for starters) hosting the back end are going to be virtual machines. Please do give suggestions to any part of the question, that you feel you have comprehensive knowledge and/or experience of. And also, point out if you feel that any part of the current set-up means an instant (or even distant) failure or if I missed a very important aspect to consider. I'm not looking for a definitive answer for how to achieve my goals, because there certainly isn't one, for I haven't provided you w/ all the required information. I'm just looking for recommendations and directions on what to look into. Also, bare in mind, that this isn't something that I have to get done quickly, to sell and let it be re-written by the new owner (which, I've been told for multiple times, is what I should aim for). I have all the time in the world and I really just want to learn doing something really high-end. Also, excuse me if my language isn't the best, I'm not a native. Anyway. Thanks in advance to anyone, who takes the time to help me out here. PS. When I do seem to come up w/ a good architecture/design for this project, I will certainly make it an open project and keep you guys up to date w/ it's development. As in what you could have told me earlier and etc. For obvious reasons the very same question got closed on SO, but could you guys still help me?.

    Read the article

  • Fluid CSS: floating column with max-width and overflow

    - by Ates Goral
    I'm using a fluid layout in the new theme that I'm working on for my blog. I often blog about code and include <pre> blocks within the posts. The float: left column for the content area has a max-width so that the column stops at a certain maximum width and can also be shrunk: +----------+ +------+ | text | | text | | | | | | | | | | | | | | | | | | | | | +----------+ +------+ max shrunk What I want is for the <pre> elements to be wider than the text column so that I can fit 80-character-wrapped code without horizontal scroll bars. But I want the <pre> elements to overflow from the content area, without affecting its fluidity: +----------+ +------+ | text | | text | | | | | +----------+--+ +------+------+ | code | | code | +----------+--+ +------+------+ | | | | +----------+ +------+ max shrunk But, max-width stops being fluid once I insert the overhanging <pre> in there: the width of the column remains at the specified max-width even when I shrink the browser beyond that width. I've reproduced the issue with this bare-minimum scenario: <div style="float: left; max-width: 460px; border: 1px solid red"> <p>Lorem ipsum dolor sit amet, consectetur adipisicing elit</p> <pre style="max-width: 700px; border: 1px solid blue"> function foo() { // Lorem ipsum dolor sit amet, consectetur adipisicing elit } </pre> <p>Lorem ipsum dolor sit amet, consectetur adipisicing elit</p> </div> I noticed that doing either of the following brings back the fluidity: Remove the <pre> (doh...) Remove the float: left The workaround I'm currently using is to insert the <pre> elements into "breaks" in the post column, so that the widths of the post segments and the <pre> segments are managed mutually exclusively: +----------+ +------+ | text | | text | +----------+ +------+ +-------------+ +-------------+ | code | | code | +-------------+ +-------------+ +----------+ +------+ +----------+ +------+ max shrunk But this forces me to insert additional closing and opening <div> elements into the post markup which I'd rather keep semantically pristine. Admittedly, I don't have a full grasp of how the box model works with floats with overflowing content, so I don't understand why the combination of float: left on the container and the <pre> inside it cripple the max-width of the container. I'm observing the same problem on Firefox/Chrome/Safari/Opera. IE6 (the crazy one) seems happy all the time. This also doesn't seem dependent on quirks/standards mode. Update I've done further testing to observe that max-width seems to get ignored when the element has a float: left. I glanced at the W3C box model chapter but couldn't immediately see an explicit mention of this behaviour. Any pointers?

    Read the article

  • Dojo extending dojo.dnd.Source, move not happening. Ideas?

    - by Soulhuntre
    Hey all, Ok... I have a simple dojo page with the bare essentials. Three UL's with some LI's in them. The idea si to allow drag-n-drop among them but if any UL goes empty due to the last item being dragged out, I will put up a message to the user to gie them some instructions. In order to do that, I wanted to extend the dojo.dnd.Source dijit and add some intelligence. It seemed easy enough. To keep things simple (I am loading Dojo from a CDN) I am simply declating my extension as opposed to doing full on module load. The declaration function is here... function declare_mockupSmartDndUl(){ dojo.require("dojo.dnd.Source"); dojo.provide("mockup.SmartDndUl"); dojo.declare("mockup.SmartDndUl", dojo.dnd.Source, { markupFactory: function(params, node){ //params._skipStartup = true; return new mockup.SmartDndUl(node, params); }, onDndDrop: function(source, nodes, copy){ console.debug('onDndDrop!'); if(this == source){ // reordering items console.debug('moving items from us'); // DO SOMETHING HERE }else{ // moving items to us console.debug('moving items to us'); // DO SOMETHING HERE } console.debug('this = ' + this ); console.debug('source = ' + source ); console.debug('nodes = ' + nodes); console.debug('copy = ' + copy); return dojo.dnd.Source.prototype.onDndDrop.call(this, source, nodes, copy); } }); } I have a init function to use this to decorate the lists... dojo.addOnLoad(function(){ declare_mockupSmartDndUl(); if(dojo.byId('list1')){ //new mockup.SmartDndUl(dojo.byId('list1')); new dojo.dnd.Source(dojo.byId('list1')); } if(dojo.byId('list2')){ new mockup.SmartDndUl(dojo.byId('list2')); //new dojo.dnd.Source(dojo.byId('list2')); } if(dojo.byId('list3')){ new mockup.SmartDndUl(dojo.byId('list3')); //new dojo.dnd.Source(dojo.byId('list3')); } }); It is fine as far as it goes, you will notice I left "list1" as a standard dojo dnd source for testing. The problem is this - list1 will happily accept items from lists 2 & 3 who will move or copy as apprriate. However lists 2 & 3 refuce to accept items from list1. It is as if the DND operation is being cancelled, but the debugger does show the dojo.dnd.Source.prototype.onDndDrop.call happening, and the paramaters do look ok to me. Now, the documentation here is really weak, so the example I took some of this from may be way out of date (I am using 1.4). Can anyone fill me in on what might be the issue with my extension dijit? Thanks!

    Read the article

  • Populate Multiple PDFs

    - by gmcalab
    I am using itextsharp to populate my PDFs. I have no issues with this. Basically what I am doing is getting the PDF and populating the fields in memory then passing back the MemoryStream to be displayed on a webpage. All this is working with a single document PDF. What I am trying to figure out now, is merging multiple PDFs into one MemoryStream. The part I cant figure out is, the documents I am populating are identical. So for example, I have a List<Person> that contains 5 persons. I want to fill out a PDF for each person and merge them all into one, in memory. Bare in mind I am going to fill out the same type of document for each person. The problem I am getting is that when I try to add a second copy of the same PDF to be filled out for the second iteration, it just overwrites the first populated PDF, since it's the same document, therefore not adding a second copy for the second Person at all. So basically if I had the 5 people, I would end up with a single page with the data of the 5th person, instead of a PDF with 5 like pages that contain the data of each person respectively. Here's some code... MemoryStream ms = ms = new MemoryStream(); PdfReader docReader = null; PdfStamper Stamper = null; List<Person> persons = new List<Person>() { new Person("Larry", "David"), new Person("Dustin", "Byfuglien"), new Person("Patrick", "Kane"), new Person("Johnathan", "Toews"), new Person("Marian", "Hossa") }; try { // Iterate thru all persons and populate a PDF for each foreach(var person in persons){ PdfCopyFields Copier = new PdfCopyFields(ms); Copier.AddDocument(GetReader("Person.pdf")); Copier.Close(); docReader = new PdfReader(ms.ToArray()); Stamper = new PdfStamper(docReader, ms); AcroFields Fields = Stamper.AcroFields; Fields.SetField("FirstName", person.FirstName); } }catch(Exception e){ // handle error }finally{ if (Stamper != null) { Stamper.Close(); } if (docReader != null) { docReader.Close(); } }

    Read the article

  • Git for Websites / post-receive / Separation of Test and Production Sites

    - by Walt W
    Hi all, I'm using Git to manage my website's source code and deployment, and currently have the test and live sites running on the same box. Following this resource http://toroid.org/ams/git-website-howto originally, I came up with the following post-receive hook script to differentiate between pushes to my live site and pushes to my test site: while read ref do #echo "Ref updated:" #echo $ref -- would print something like example at top of file result=`echo $ref | gawk -F' ' '{ print $3 }'` if [ $result != "" ]; then echo "Branch found: " echo $result case $result in refs/heads/master ) git --work-tree=c:/temp/BLAH checkout -f master echo "Updated master" ;; refs/heads/testbranch ) git --work-tree=c:/temp/BLAH2 checkout -f testbranch echo "Updated testbranch" ;; * ) echo "No update known for $result" ;; esac fi done echo "Post-receive updates complete" However, I have doubts that this is actually safe :) I'm by no means a Git expert, but I am guessing that Git probably keeps track of the current checked-out branch head, and this approach probably has the potential to confuse it to no end. So a few questions: IS this safe? Would a better approach be to have my base repository be the test site repository (with corresponding working directory), and then have that repository push changes to a new live site repository, which has a corresponding working directory to the live site base? This would also allow me to move the production to a different server and keep the deployment chain intact. Is there something I'm missing? Is there a different, clean way to differentiate between test and production deployments when using Git for managing websites? As an additional note in light of Vi's answer, is there a good way to do this that would handle deletions without mucking with the file system much? Thank you, -Walt PS - The script I came up with for the multiple repos (and am using unless I hear better) is as follows: sitename=`basename \`pwd\`` while read ref do #echo "Ref updated:" #echo $ref -- would print something like example at top of file result=`echo $ref | gawk -F' ' '{ print $3 }'` if [ $result != "" ]; then echo "Branch found: " echo $result case $result in refs/heads/master ) git checkout -q -f master if [ $? -eq 0 ]; then echo "Test Site checked out properly" else echo "Failed to checkout test site!" fi ;; refs/heads/live-site ) git push -q ../Live/$sitename live-site:master if [ $? -eq 0 ]; then echo "Live Site received updates properly" else echo "Failed to push updates to Live Site" fi ;; * ) echo "No update known for $result" ;; esac fi done echo "Post-receive updates complete" And then the repo in ../Live/$sitename (these are "bare" repos with working trees added after init) has the basic post-receive: git checkout -f if [ $? -eq 0 ]; then echo "Live site `basename \`pwd\`` checked out successfully" else echo "Live site failed to checkout" fi

    Read the article

  • Login Script for PostGreSQL and PHP not working =[

    - by MrEnder
    Ok I'm quite new at logins what not so bare with me here lol but I gota learn so don't discourage me. I tried this so far <?php $error = ""; $conn = pg_connect("host=localhost dbname=brittains_db user=brittains password=XXXX" ); $sql = "SELECT * FROM logins"; $result = pg_query($conn, $sql); if($_SERVER["REQUEST_METHOD"] == "GET") { $userName=""; $password=""; } else if($_SERVER["REQUEST_METHOD"] == "POST") { $userName=trim($_POST["userNameLogin"]); $password=trim($_POST["passwordLogin"]); if(pg_fetch_result($results, $userName, "userName")==true && pg_fetch_result($results, $password, "userName")==true) { setcookie("userIDforDV", $userName, time()+43200); } else { $error = "Your username and or password is incorrect"; } } $userName = $_COOKIE['userIDforDV']; if(isset($userName) && $userName!="") { echo "Welcome " . $userName; } echo $error; ?> <form> <table> <tr> <td class="signupTd"> User Name:&nbsp; </td> <td> <input type="text" name="userNameLogin" value="" size="20" /> </td> </tr> <tr> <td class="signupTd"> Password:&nbsp; </td> <td> <input type="password" name="passwordLogin" value="" size="20" /> </td> </tr> <tr> <td class="signupTd" colspan="2"> <input type="submit" name="submit" value="Submit"/> </td> </tr> </table> </form> that was the idea I came up with... but its prolly a really bad idea and it doesn't work... how might I go about this properly? I need really detailed descriptions please. Thanks a tun Shelby

    Read the article

< Previous Page | 17 18 19 20 21 22 23 24 25  | Next Page >