[MUD-Dev] Version Control (was: DBs and Events)

coder at ibm.net coder at ibm.net
Tue Feb 17 12:42:21 CET 1998


On 16/02/98 at 03:37 PM, Vadim Tkachenko <vadimt at 4cs.com> said:
>coder at ibm.net wrote:

>[skipped]
>> I am becoming a fast fan of journalling versioning file systems.  These
>> problems just disappear.

>Can you please elaborate on that? 

Journalling filesystems (loosely speaking) run the filesystem as a
journalling database.  The result is that should there be an ungraceful
shutdown of the filesystem, istead of corruption the filesystem is instead
rolled back to the last compleat IO (comitted transaction).  The result is
that journalling filesystemsare somewhat slower than "straight"
filesystem, but recover (fsck time) much more quickly, and with far less
(close to none) risk of significant loss.

Versioning filesystems are best known with VAX/VMS.  A versioning
filsystem automatically generates and keeps track of previous versions of
files.  In VMS's case, a filename was specced as
"<name>.<extension>;<version>"  Opening a file by its name.extension would
default to the latest version (you could also spec a specific version to
operate on).  Any write to that file would create a new file with that
change with an incremented version number, leaving the original file
untouched.  Of course you could also purge previous versions, set version
hostory limits, set certain files as having no history etc.

>Right now I'm trying to choose the
>right one for UNIX project[s], so far figured out how to use RCS, looking
>for information on CVS, know about anything else?

RCS, CVS, SCCS and company are version control systems.  They have nothing
to do with filesystems per se -- they operate at the file or directory
level.  I use RCS -- I haven't looked into CVS (tho I suppose I should). 
Comparitively RCS is a superset of SCCS, and is quite simple.

Note a versioning filesystem would essentially do RCS int he background
without your ever having to touch it.  cf ClearCase.

>RCS is good enough, the only thing I haven't figured out yet is how to
>allow multiple users to edit and compile the code with a minimum overhead
>- the kind of people I'm dealing with is having difficulties setting
>CLASSPATH...

Spec the problem out a little more and I may be able to help.

--
J C Lawrence                               Internet: claw at null.net
----------(*)                              Internet: coder at ibm.net
...Honourary Member of Clan McFud -- Teamer's Avenging Monolith...




More information about the mud-dev-archive mailing list