Hacker News Reader

Comment by gchpaco | Hacker News Reader

A lot of the problems SVN ran into was trying to use cutting edge technology and standards before they were really ripe. WebDAV, for example, or the endless, endless iterations of Berkeley DB corruption. It is, in a lot of ways, an excellent illustration of the dangers of trying to solve the wrong problem.

View on HN

What's innovative about WebDAV ? It's basically a design by committee standard pushed by Rational to web enable clearcase; it's a very verbose protocol and there was an entry in Sussman's blog about replacing WebDAV with a better wire protocol like Mercurial's. Getting WebDAV into SVN was basically Greg Stein's idea - he had worked at Microsoft and had observed them transitioning over to WebDAV in a big way and started pushing for its usage in SVN as well - a move that was not well thought out. WebDAV is not very popular even with some of the SVN hackers.

Berkeley DB was certainly old when it was adopted by the SVN project. It's not like the NFS issues with BDB were unknown.

I never said WebDAV was innovative, just that it was brand new and full of problems. And as for BDB, IIRC they were having issues with the database getting wedged constantly even when not using NFS. But this was long after I had stopped paying a great deal of attention to the project.

"he had worked at Microsoft and had observed them transitioning over to WebDAV in a big way and started pushing for its usage in SVN as well"

That's why I am suspicious anyone who ever worked for Microsoft, specially when you talk about open-source. You never know when they will send a saboteur.

Linus was making fun of SVN in his talk ( http://www.youtube.com/watch?v=4XpnKHJAok8 ). One of his point was, that SVN chose the wrong way right from the start and there is no way to have "CVS done right" because CVS is hopeless by design.

CVS may have had serious flaws, but I find Linus' original revision control practices to be less than optimal:

"The Linux kernel source code used to be maintained without the help of an automated source code management system, mostly because of Linus Torvalds' dislike of centralized SCM systems.

In 2002, Linux kernel development switched to BitKeeper, a SCM system which satisfied Linus Torvalds' technical requirements."

Nearly a decade without a revision control system?

http://en.wikipedia.org/wiki/Linux_kernel#Revision_control

In Linus' git talk, he says that he finds creating different tarballs is much better version control than using CVS.

If you compare granularity of tracking and utility of historical version control across different projects of that era, such as FreeBSD vs. Linux, I think you'll find that Linus' position is clearly false.

Reply on news.ycombinator.com