Re: [Monotone-devel] Proposition: hooks to set permissions on branch names

[Derek Scherger]

Richard Levitte - VMS Whacker wrote:
> Hi guys,
>
> there are times when one might want to assign extra permissions to
> some people for some branches rather than others.  Right now, it seems
> like such possibilities are missing (I was looking at the hook
> get_manifest_cert_trust, but found it had a very different purpose),

I believe the purpose of the trust hooks is more about controlling your view of what's in 
your database rather than controlling what gets into your database. i.e. if there's some 
stuff in your database you don't like and you don't trust it you don't see it and it 
essentially appears as if it isn't in your database for things like update, merge, etc.

> so I'm proposing the following hooks:
>
> get_read_permitted (branchname, identity)
> get_write_permitted (branchname, identity)
> get_anonymous_read_permitted (branchname)
>
> They would work just like the get_netsync_*_permitted hooks, but be
> called anytime a read or write operation is to be performed, based on
> the branch cert found with the object that was just fetched or the
> object that is about to be written.

Here's the thing, you have your database, I have mine, you have your hook settings, I have 
mine. You can set your hooks to control what you can commit to your database, but I can 
set mine  however I like, and since it's *my* database, presumably I'll want to be able to 
commit to it. ;)

> To tell the truth, I see no reason why the get_netsync_*_permitted
> hooks would stick around if the above is added, as they would become
> some kind of lesser duplicates...  And considering Graydon's thoughts
> about removing the concept of collections altogether, this might be a
> good idea either way.
>
> I haven't looked at that part of the code yet, so I've no clue how
> hard this would be to implement.  I also have no clue how this would
> affect performance.
>
> If there is a way to do what I'm looking for with currently available
> hooks, I'd very much like to be informed (clued in, really :-)).

I think the *_netsync_permitted hooks let you control who can put stuff into your 
database. Although now that I think about it, I'm not sure they will prevent *you* from 
pulling in something by someone you don't want to allow to write to a particular branch.

Hmm... is this where you were going with your proposal? i.e. controlling which versions 
arrive in your database when doing a netsync based on their authors rather than on who is 
doing the netsync? This seems like it might be an interesting idea but I'm not really 
sure, this is all pretty new to me too.

--
Cheers,
Derek