"hg log" for remote repositories?

[Matt Mackall]

On Thu, Sep 13, 2007 at 12:41:24PM -0700, Jens Alfke wrote:
> 
> On 13 Sep '07, at 10:32 AM, Matt Mackall wrote:
> 
> >I suppose "ssh user at server hg log" or "firefox http://server" is  
> >cheating?
> 
> This is for use in a GUI app that will display the files and revision  
> history. I'd like it to be able to work [at least partially] on remote  
> repositories as well as local ones.
> 
> The ssh command approach is a good idea, but of course it only works  
> with ssh-based URLs, not HTTP.
> 
> >Sure. But incoming is doing exactly the same thing as pull over the
> >wire. Which is why incoming -vp, etc., works.
> 
> Does that mean that "hg incoming" actually retrieves (and then  
> ignores) all the contents of the revisions? As an experiment I ran "hg  
> incoming http://selenic.com/repo/hg-stable" from an empty repository,  
> and it took nearly a minute to run. I don't know how many bytes of  
> data it pulled over the network, though.
> 
> >We could do log over the wire,
> >but it would be a) very complicated or b) missing lots of features  
> >or c)
> >*SLOW*.
> 
> Hm. I don't know the implementation, but remote "log" seems like a  
> subset of the "incoming"-from-empty-repo trick, only skipping the  
> bodies of the change-sets.

Until you want -p support or basically anything non-trivial. In other
words: b) missing lots of features. And then people will want those
features and we'll be back at (a).

The whole point of a distributed system is to not do slow operations
on remote repos.

-- 
Mathematics is the supreme nostalgia of our time.