Hacker News Reader

Comment by nickpsecurity | Hacker News Reader

Interesting. You're the first person I've seen here that used it. It was the only one that meets most of Wheeler's requirements:

http://www.dwheeler.com/essays/scm-security.html

So, was it a good tool compared to subversion or centralized git? Work fast and reliably enough? Not get in the way too much outside authorization checks? Just getting feedback on field use here.

View on HN

I don't think I know enough about subversion or centralised git. Aegis was great for the first few years. For the main project, there was a perception that it was slow and unreliable. But a lot of us loved the transactional features and preservation of the timelines of comments for code review. Its management of test scripts was very very valuable.

I'm not sure about your authorisation checks question. Maybe Aegis did get in the way.

An afterthought about speed and reliability - Aegis is almost too configurable. For instance, it relies on a 'file history tool' per project; that can be SCCS, RCS, FHist, or anything that you can tell Aegis how to use. We used the somewhat obscure FHist. It has the advantage that it was written by the same person as Aegis, and so is extensively tested with it.

Our project had the main repository sitting on an NFS server, and in later years around a dozen working on one or two changes at any one time, with those stored on each engineer's Linux workstation. That was slow and sometimes fragile, but we never really explored other confiration options.

We accumulated hundreds of unit test scripts which had to pass on each integration. In retrospect we should have worked harder to retire or refactor the oldest ones. Having fewer or faster tests would have saved time and angst.

Aegis' implementation is very dependent on the Unix security model. We thought about doing something to make it work on Windows, but quickly got discouraged.

I don't know much about security, but I'd guess you'd be able to discover many.

Sadly, Peter Miller, the originator and I guess owner of Aegis, died a few years ago. User and developer mailing lists were on the web last time I checked.

Thanks for the feedback. Interesting. The UNIX dependence was one of my main gripes. I had some workarounds but figured a clean-slate model or retrofit of cross-platform SCM was better. I still wanted to eventually meet and get insight from the author on such a project. Too bad he died. (sighs)

Back to clean-slate or retrofit of alternative as I figure Aegis might be too complex to use for such a project if main mind behind it isn't available. Still, they get credit for doing a lot of things right.

Yes. I might suggest you give one of the tutorials a try. He as several, depending on which tools you're more comfortable with.

Creating a small project seemed to be quite a bit of work, and I think Peter realised this and put effort into the examples.

You might also like his PDF on recursive Make, http://aegis.sourceforge.net/auug97.pdf.

Reply on news.ycombinator.com