As Mark pointed out, I'm investigating the technical details on how we should migrate from TeamWare to a more recent SCM.

At the moment I'm busy gathering data about Mercurial to determine the impact it will have on disk usage and time it takes to bringover (the TeamWare equivalent of the Mercurial clone command).

I am also working with Kelly on how we should go about an migrate all information in TeamWare and SCCS to Mercurial. Right now I'm not worrying about non-technical details, e.g. if the SCCS comments are encumbered, i.e. if they contain information which can not be open sourced. I've written a simple Perl script, reusing what Steve have done for OpenSolaris, which allows me to iterate over all SCCS revisions, and do a "hg add" of revision 1.1 and then "hg commit" on all the following revisions.

Doing a simple migration like that is not a problem, but we'd like to group each TeamWare putback into a change set, and that requires some magic. One problem is that TeamWare records time in local time of the engineer who ran the "sccs delget" command, so when you have many engineers working on code that gets integrated at the same time, we can not use the timestamp to group the entries.

Another problem is that the Java SE source have had very relaxed requirements on the SCCS comments, especially if you compare it to Solaris. Lately it has gotten a lot better, but it is far from as stringent as what you find in the ON RTI process.

You can find other random notes about Mercurial on my other blog.

[Technorati Tags: Java SCM Mercurial ]