I need to decide on a method for storing a subset of the content in a SharePoint site, so that when I delete and recreate certain lists as part of a feature activation, I can re-insert all of this content back where it should belong. I have an idea myself, but I don't know if it's the only method and more importantly, the right method.

My client has me creating a SharePoint system for them to communicate with their clients. The business process has maybe 5 stages in it (maybe it's more, I don't even know because they don't tell me everything), and the current system I've written over the past months is maybe 2 stages through. This meets our deadline of completing those systems by Monday next week... but at that point my client is planning on making the site live from that point.

In effect, their work with their clients will be running parallel with my work for them. As I complete my own work on a separate test server, I'll push each following stage of the process onto the live server. Scheduled downtimes during non-business times (like a weekend) will be available for me to perform these pushes. Keeping pace so that my development is faster than the actual business process is my own problem and off-topic... so let's get back to the problem I stated at the start of this post.

In this system, we have sets of features which will create lists for their associated content types and field types when activated, and delete these lists when the feature is deactivated. Most updates don't need to deactivate and reactivate these features, such as workflow changes, custom actions, custom forms, and similar ilk. But there are some parts which do require this. On my test server, it's okay for me to obliterate lists, but once the site is live and there's real correspondence data, it's absolutely unacceptable to do this. So when I need to implement a new change in functionality, I need to be able to store the currently present data in several lists, deactivate the feature, reactivate the feature, and restore all of this data.

Perhaps I have hoist myself by my own petard with the feature system I implemented. Unfortunately, the necessity to later on make several of these "project sites" meant I had to do a lot of my code with the concept of "Can be deployed repeatedly" in mind.

My current plan is to run through lists and libraries which will be affected by the particular feature that is to be reset. Files and all of their versions will be saved in a directory on the server. Then, a set of text files will be used to store all of the important field values for the items. This includes a lot of cross-list reference lookups that will need to be maintained, but that's simple enough. Then, I deactivate the feature, deploy the new solution, and reactivate the feature. We upload all of the files in the order specified by their versions and update them with the stored fields for those versions, so that we retain the version structure. As each one is first uploaded, the new ID is picked out, and all relevant lookups in the rest of the files are updated (in some manner that I make sure I don't re-update it later with an incorrect value, of course). After that, we run through all the rest of the items in the order most conducive to keeping the relational data correct. This roughly summarizes what my current plan is. To my advantage, there are no long running workflows in the system that will be affected by this, so there's nothing I will have to worry about making sure nothing is "still running" when I do this stuff.

I don't really know all the cons of this approach... I can imagine they're quite hefty. But I'm unsure what other choices I even have, and my searches haven't turned up anything. Is there anyone who can think of a better idea? Or will anyone just tell me that I really have no other choice? Thanks in advance!