Very yes. Though unix has extensionless files, the web and MIME are defacto
using suffixes for file type id.
At 06:07 AM 2/15/00 -0800, Paul Sander wrote:
>>--- Forwarded mail from [EMAIL PROTECTED]
>
>>Let's take a break from the regularly scheduled passive-agressive
>>flamefest and think about an imaginary Versioning System (iVS)
>>that meets the features that are being talked about. Here's the
>>first thought off the top of my head:
>
>>1) Use some kind of MIME file-type mapping to control settings
>>for locking/concurrency (so that your source files are controlled
>>concurrently but images use hard locks)
>
>Yes!
>
>>2) Use similar by-file-type mappings to controll the diff and
>>sync algorithms used (standard text diff for ascii or unicode,
>>something like Xdelta for your PNGs)
>
>Oh, yessssss....
>
>Ideally these file type mappings would be controlled by type managers that
>specify all of the priimitive operations that can be done to each type of
>file.
>
>A versionable mapping of source containers (RCS files) to working copies
>(i.e. a directory management mechanism) is also needed. This will allow
>renames without the awkward (at best) remove/add workaround, and replace
>the broken modules database implementation. Note that this has profound
>effects on how repository locks are done, so the current mechanism would
>have to be replaced if it is to scale well.
>
>And more hooks: pre- and post-operation, client- and server-side and
>ideally per-file, per-directory, per-command. And the interfaces should
>be more scalable than the current *info interfaces.
>
>>Such a system could be configured to behave no differently than
>>the current CVS, or it could be configured to always require
>>hard locks, or something in between.
>
>I believe it's possible to reconcile the user models, but the repositories
>would be somewhat different. Forward compatibility (from CVS to the new
>system) should be possible without closing off a return trip should it be
>necessary for some reason (at least, up until the first file rename is done).
>
>>I suppose you could set the mappings by project subdirectory too,
>>in case you wanted specialized control somewhere.
>
>Is there any reason why a type manager could not manage directories and
>their contents?
>
>>Ok, everybody tell me why this can't be done or why I'm talking
>>to the wrong list.
>
>>[ducks and runs]
>
>Stand proud, my friend! But break out the armor and asbestos underwear
>just in case...
>
>>--- End of forwarded message from [EMAIL PROTECTED]
>
