Re: Branches as signed sequences Justin Mason (
Thu, 13 Apr 2000 12:23:49 +0100

Kai Henningsen said:

> Oh, and (5) security from non-owner non-root users who might have file
> level access to the repository so they can get at parts of same. A client-
> server and/or set-userid scheme would solve this. Or you could handle that
> by putting the entire repository under a key that only the owner has - but
> the repository server needs access to that key, and the other users must
> not have access, and if your OS file-level protection doesn't give you
> that, you lose.

Yep, I'd give a thumbs-up to a client-server scheme too, (a) to deal with this issue and (b) to make the system easier to use on a network of many client machines talking to 1 repository server.

To explain (b): the idea would be that a client machine would not have to support NFS to access the repository, and would hopefully be easier to port to new platforms -- so that users of those platforms would be able to use this DCMS even if they couldn't get all the server-side repository logic ported.

BTW I'm afraid this is yet another Perforce feature ;)

I get the impression a lot of folks here are unfamiliar with p4, and IMHO it would be worthwhile taking a look at it before it's too late to nick design ideas; it has a lot of features that are a good next step from the CVS model, and it's learnt from several other systems' mistakes already. Here's a good architectural overview:

Regarding directories:

>This differs strongly from the view expressed in the tigris design. In
>general, I see no reason to capture directories as long as paths are fully

I agree.

Justin Mason       Work: <>
                   Personal:     <>

"It's true that some sharks get cancer. I said this in my book."
	 	   -- William Lane, author of _Sharks Don't Get Cancer_