These discussion always remind me of a talk I had with an elderly
German man who was nearing retirement from a manufacturing company I
worked for. They made a lot of stuff that used a lot of parts in
each unit. They started with a part numbering sachem that "had
meaning" to the numbers. xxx-yyy-zzz type thing where the x's were
the product family, the y's where something else and the z's
something else. This gentleman had been in charge of assigning these
numbers for over 30 years. He said that one day, after many reworks
of the meanings, many heated meetings, and after years of his
requests the management finally allow him to just start assigning
sequential numbers. He said the bottom line is that no matter how
hard you try to attach meaning to a number it never works in the long
run. People assign them against the standard, people forget, the
standard wont adapt to the new conditions; it's always something. He
said what matters is that you have a place to write down what you
mean and you have some way to look up the number from what you mean;
if you REALLY need to.
I think Subversion has done a nice job of encapsulating this
wisdom. Maybe you can argue that a file system directory table is
not the best place to 'write down what you mean' but, my considerably
less experience and that gentleman's years of experience continues to
indicate to me that trying to assign meaning to numbers is a hopeless endeavor.
If you need to track down some specific state of not-so-broken code
the tools ("svn log" and / or "svn blame") are there. Once you find
what you want "tag it" by 'writing down a meaningful name' in the
directory structure ("svn copy".) It is really pretty straight
forward. The other versioning systems are not that much different.
(I have use a lot of them.) They just write the meaningful name in
some file or database somewhere. (That you generally need to do some
report or look up to see those 'meaningful' names that you assigned;
though you can't remember them any more!) I have considered it
rather effective of the SVN community to use a native structure for
storing/viewing this information rather than making up yet an another
storage method, that you need to get reports to read. After all is
not one of the biggest ideas behind version control to help reuse
stuff? SVN Copy is reused as a tag tool. The directory structure
and all your favorite OS tools for searching, sorting, hiding,
whatever, are reused to manage the list of 'meaningful names.'
At 23:23 2005/08/14 +0100, Thomas Beale wrote:
>Monks, Peter wrote:
>>G'day again,
>>
>>>often you don't know in advance that you need to roll-back to some
>>>particular revision...e.g. your developers might have introduced
>>>some inconsistencies into the codebase without realising it for a
>>>few revisions....
>>
>>Don't "svn log" and / or "svn blame" give you this information?
>>
>>>No it doesn't. It just doesn't provide enough support for doing
>>>tags, releases etc; all users are in the position of having to
>>>implement something extra, meaning somehting unique and different
>>>at each site,
>>
>>I must be missing something again - if (as I think you've suggested in
>>previous posts) there's a universally applicable CM methodology, why
>>would it need to be independently and uniquely implemented at each and
>>every site? Why not just implement it once (using the various
>>extension points Subversion provides) and reuse it (or even better,
>>share it with the wider Subversion user community so that no one else
>>has to implement it either)?
>
>yes, that would be nice wouldn't it...;-)
>
>- thomas beale
>
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
>For additional commands, e-mail: users-help@subversion.tigris.org
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Tue Aug 16 16:15:00 2005