Re: [Monotone-devel] private key needed for sync?

[Zbynek Winkler]

Brian Campbell wrote:

> You need to have a key in the server so the client can make sure it's  
> connecting to the right server. Otherwise, someone could set up a  
> bogus server so when you sync to it, all of your private code is sent  
> to the malicious server.

That sounds reasonable. But I am confused by the tutorial and/or 
monotone behavior :(. Why does it use something called "private key" for 
server identification? I thought that "private key" is for 
identification of the commiter? I got this feeling from the tutorial. It 
didn't say anything about "server identification" (or I've missed it).

Also the error message says it is looking for "signing key" and not for 
server identification key...

> As far as one database per project, you can either do that on the  
> server, in which case as you say you would have to run separate  
> server instances on separate ports, or you can have the server just  
> have one big database with all of the projects, and then have  
> individual databases for the individual projects that sync with the  
> server periodically or whenever people feel like it. If the reason  
> for having per-project databases is that you want people to be able  
> to access their own database, then this is a better idea, as it means  
> you don't need to stop the server to work on the individual database.

Well, I just get the felling that the 
"one-huge-file-with-every-revision-of-every-project" does not scale very 
well. BTW: how does monotone handle multiple connections to the netsync 
server? Can multiple users sync at the same time or could one busy 
project stall all the others?

> By the way, it looks like you've named your collection marocode. It  
> would probably be better to follow the reverse domain name naming  
> convention, so it would be cz.matfyz.zw.marocode or cz.cuni.marocode  
> or whatever domain it is you want to use. We discussed possibly  
> moving to a URL like convention, like zw.matfuz.cz/marocode, but that  
> hasn't happend yet because the slash conflicts with the slash used in  
> the selector syntax.

I felt it would be usability deficiency to force users to always specify 
the whole url when switching to another branch, so I went with shorter 
names.

I think we'd need some form of short/long branch names or some user 
aliases for the branch names or a project name (the long stuff) and a 
branch name within the project. The goal would be to specify the long 
stuff with the url only once...

Zbynek

--
http://zw.matfyz.cz/     http://robotika.cz/
Faculty of Mathematics and Physics, Charles University, Prague, Czech Republic