Keeping History

Back in mid 2005 I worked on a project for customer wherein I was generating "Froogle Feed" data for the customer's e-commerce sites. For some reason, the project never went live. Suddenly, a request shows up from the customer mentioning "Froogle Project" and asking for a couple of features.

I have no memory of how I accomplished this project in the first place. I don't know if the customer does either. I just have a couple of features in an email project and zero context. Does the customer remember the experimental work we did a couple of years ago? Does he remember the UI?

The big problem is that aside from a couple of emails and basecamp messages, and a couple of generated 'feed' files, I have nothing for this project. Looking through source code history returns nothing. Looking through the current site (an old Zope 2 based site, built out of page templates, python scripts, and SQL Methods) yields nothing either. Unfortunately, there's not a lot of history in the site. And if the feature in question never went live, there's (of course) not going to be ANY history.

I dug through a big folder of printouts and such collected over the eight-plus years of history with this customer. Nothing.

I finally turned to notepads and notebooks, and did manage to find a couple of notes in a couple of different places. And digging through some basecamp messages yielded a mention of a 'froogle.py' file in a development instance that was not put into source control (CVS at the time). That development server was killed off long ago. It may have been hobbling around, but after moving offices twice in the last year or so, the number of internal servers is nearly zero. Long unused development servers tend not to get plugged back in.

Some lessons learned:

1. Always put a date stamp on handwritten notes. This is now a habit for me, but it apparently wasn't a habit back in 2005. I did find some dates on some pages and used that to backtrack to when this work was initially done (using date stamps on Basecamp and email messages). A side lesson to this is to always include the year when date stamping.


2. Always use source control, even for unfinished projects. If this one file had been checked into CVS, I would be a bit less stressed right now. I'm far less familiar with this customer's site now than I was then and trying to remember how I built multiple Froogle files and put them into a Zip file would be great knowledge right now, even if the format has changed. Since moving to Git, it's easier to keep dead branches in the central repository. CVS made branching so painful that we never would start a 'froogle-feed' branch. Even if that branch never goes live, it's damn useful to have.


3. Don't trust development servers. They may not be around forever, and odd little unfinished files or projects can get lost very easily over time. May not seem like a big deal, but sometimes these little projects come roaring back to life after five years. We now do most of our development on our desktop machines. When we get a new machine, all of our files roll over. Central servers work if it's a central common copy (like a master account server). I've found that development servers just lose luster over time, and tend to get very far behind in terms of hardware and core OS performance. Plus they tend not to have the kind of machine/account migration tools that desktop OS's like Mac OS X have.