Anonymous access to repository via CVS client?

[Greg Noel <Gr...@tigris.org>]

[Possible duplicate---I sent this yesterday, but it never showed up on 
the mailing list]

I originally submitted this as an issue and was told I should post it 
to this mailing list instead.  My background is that I am a heavy CVS 
user, with lots of scripts to do various things, but I'd rather be a 
SVN user since I believe it would solve a number of problems I have 
with CVS.

But I don't believe that Subversion will be able to make serious 
inroads on the CVS mindshare until it can transparently provide 
anonymous read-only access to a repository via a standard CVS client.  
As part of the Subversion package, I believe you need to provide a 
simple CVS server that can check out a "tag" and provide updates for it.

As a rule, the users of a project greatly outnumber the developers (at 
least, one hopes that's the case).  Moreover, the developers tend to be 
a tight-knit community, so they can coordinate a switch to a new 
utility.  If the impact of moving to Subversion only means that the 
database needs to be converted and the developers are inconvenienced a 
little bit, it becomes much easier to sell, since the user community 
can change over at their own convenience.

Note that I'm _not_ talking about a full implementation of the CVS 
protocol.  The only portions that need to be implemented are the 
portions related to checkout and update.  Nothing that changes the 
repository.  It'd be good to support diffs and maybe status, but that 
needn't be done initially; it isn't really needed to support a 
conversion.  Nor am I implying that it needs to be implemented in a 
compiled language: it doesn't need to be terribly efficient, since it's 
only needed while the user population is converting, so a language like 
Python or Perl would make a lot of sense.

For example, look at SourceForge.  From some comments by their staff, 
they would very much like to switch to Subversion, but the magnitude of 
converting all the anonymous pserver-based read-only access is too 
daunting.  If they could transition projects and only impact the 
developers, they would be much more willing to start converting.

Moreover, providing this access makes a number of things simpler.  
Tools like cvsweb and their ilk would "just work" and wouldn't need to 
be modified immediately.  And getting the initial source download onto 
architectures where there isn't a binary starter kit (like my PPC-based 
Linux system) is much easier.

Thank you for your attention to this.

-- Greg Noel, retired UNIX guru


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: Anonymous access to repository via CVS client?

[kf...@collab.net]

Greg Hudson <gh...@MIT.EDU> writes:
>   * I think this should have a home in the issues db.

If someone starts it, they should feel free to file an issue to track
it.  But in the absence of concrete action, an issue would just be
clutter (let's not use the issues db as a general brainstorm bucket).

Just my $0.02,
-K

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: Anonymous access to repository via CVS client?

[Greg Noel <Gr...@tigris.org>]

Sorry for the delayed response: I'm one of those people in Southern 
California that's been dodging wildfires this past week and it's taken 
me this long to catch up.

On Sunday, Oct 26, 2003, at 23:37 US/Pacific, Greg Hudson wrote:

> This sounds like a good idea, and I might look into it myself ...

Thanks; I appreciate the vote of confidence.

>   * I think this should have a home in the issues db.

Yes, that's why I put it there initially.  I wouldn't object if you 
re-opened the ticket...

>   * It shouldn't be a 1.0 requirement, nice as it would be to have.

I think it's regrettable, but I agree.  From what I read in the mailing 
list, you're too close to shipment to add an additional requirement.

>   * There might be some snags.  CVS does not have a reputation for
>     great protocol quality.

I'm sure there will be lots of snags, but even a horribly inefficient 
implementation (viz., check out a copy from SVN, check it in to RCS, 
then use a real CVS server from the RCS files) would be sufficient 
initially.

> We do have source tarballs for just this reason.

Yes, but those are only produced for release points.  If the tarball 
isn't sufficient to provide a working program, then it could be weeks 
before a fixed version is available.  Being able to check out the 
sources between tarballs would ease this situation.

It's a small point, anyway; the essence of this issue is to reduce the 
trauma of converting to SVN by allowing a transition period instead of 
a massive Flag Day.  (I was there at the original Flag Day; it's an 
experience I'd rather not repeat.)

Hope this helps,
-- Greg Noel, retired UNIX guru


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: Anonymous access to repository via CVS client?

[Greg Hudson <gh...@MIT.EDU>]

This sounds like a good idea, and I might look into it myself if I can
finish some other changes while I'm still being productive.  (No one
else should let that stop them.)  So:

  * I think this should have a home in the issues db.
  * It shouldn't be a 1.0 requirement, nice as it would be to have.
  * There might be some snags.  CVS does not have a reputation for great
    protocol quality.

> And getting the initial source download onto architectures where there
> isn't a binary starter kit (...) is much easier.

We do have source tarballs for just this reason.


---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org