[Monotone-devel] bidirectional CVS gateway

[Christof Petig]

[intentionally breaking deep the mail threading]

graydon hoare schrieb:
> and we've been over this before. it usually boils down to some
> variation (synchronous or asynchronous) of having a "branch owner" --
> person, robot, discriminated server -- issue a special linear cert.
> this unfortunately centralizes a failure case, and is not robust
> against backup/restore and multiple-issue errors on the case of the
> owner.
>
> unless oren has some new idea about how to establish a stable, robust
> linear subgraph, I propose we just let the whole topic drop for a
> while.  we've been running around in circles all week, and not come
> up with much new since the last time this was raised 6 months ago.
> we're all getting bitchy, and there's not much gained by making the
> mood hostile. I already made enough bad blood by saying that phonetic
> hashes sounded "crazy"; I'm sorry for having said so already, but
> it's done. let's just focus on real bugs and easily-agreed-on
> improvements for a while.

I'm sorry I can't drop this topic. Switching to monotone to do real 
_work_ would be dramatically easier if we had a monotone CVS 
synchronization.

I said I would code it (given that you think it possible) and already 
looked deep into the CVS server protocol. To make a sync monotone->CVS 
you need the concept of a HEAD revision (and to make the sync 
_efficient_ you should not choose a new HEAD which is not a descendant 
of an old HEAD). So certificates which indicate which DAG branch is a 
new CVS branch and which one continues an existing CVS branch would be 
needed as a prerequisite to synchronization (I might abuse timestamps to 
get a consistent though unstable notation).

All I need is a cert saying "take me as the main branch" which is trusty 
enough to get considered. Multiple credible (brach-)HEAD-certs must get 
resolved before sync can take place (like update, which says the branch 
is unmerged and exits)

------------------

Apart from that: Do you have any design ideas on how to construct a 
bidirectional CVS gateway? Another (less elegant) solution is to put a 
CVS checkout into monotone (with CVS/Root,CVS/Repository etc). You get 
free syncing by updating and then committing into the other VCS.
[This might satisfy my needs but not my taste]

IMHO the main problem is to identify corresponding versions in monotone 
and CVS [timestamp and changelog might be your only guide] without 
downloading every revision of every file (to get the SHA1 ID). cvs log 
(or even less e.g. Entry/Unchanged) should be the main interface to 
synchronize. But to use "Entry/Unchanged" we need to store the CVS file 
version in monotone. Any ideas?

Pull idea: Iterate over every file, Iterate over every branch, get the 
head revision, get the changelog and timestamp of the head, try to 
identify monotone version, unless you find the last pulled version, 
create new MT revisions as needed.

Push idea: determine (how???, cvs log on each file?) which revisions are 
present in CVS and which are missing, determine which of these modify 
which branch (in order) and perhaps (option?) create additional branches 
on the fly (you know, I'm kidding if you know the CVS protocol, it does 
not fly for branches). Sounds like there should be an optimized version 
which does only determine and commit new changes to the head of a 
specific branch (looks less network intensive). This sounds more like 
merkle trees (I don't know their theory) might come in handy.

   Christof

signature.asc
Description: OpenPGP digital signature