[Monotone-devel] [RFC] monotone URIs

[Thomas Keller]

Hi all!

I asked around on IRC if the monotoners ever thought about 
monotone-style URIs (so this issue might already have been discussed to 
death before my time, if this is so, please ignore this mail), and 
apparently the answer was yes, with some "this is needed" addition.

Now for what are those URIs useful? Mainly they're convenient for 
newbies and general communication. You don't have to give somebody a 
list of configuration details, just one URL which (nearly) contains 
everything and you're done. They could also make our interface a little 
cleaner as it is now and even introduce new features, more on this in 
just a second.
Commands which particularily benefit from URIs are push, pull, sync and 
(the upcoming) clone. So how could such an URI look like? A simple form 
like this:

 $ mtn pull mtn://<host>[:<port>]/<rev-selector>

Note that there is no path to the monotone db file specified, as there 
is (to my knowledge) no easy way to run serve for two databases on one 
and the same port anyways (never played with usher, though).

'host' is a normal hostname. Note that I didn't include the known 
username phrase user@ optionally before the host; this is mainly because 
I believe email addresses which we use as key names / user names would 
conflict with the @ sign in there. So either we "rewrite" the key here 
somehow (like email:address.com) or leave that out completly.

'port' is optional and defaults to monotone's default port 4691

'rev-selector' is a revision selector which is already known for other 
commands. Revision selectors should be able combineable (like 
a:foo/t:some-tag) and in order to support include/exclude patterns which 
are currently separate options in push/pull/sync, the introduction to 
negated selectors is needed. So, e.g. to pull all branches beside foo 
and bar one would write

 $ mtn pull mtn://my.host.net/b:*/!b:foo/!b:bar

Since selectors select revisions and not branches by default, its 
obvious that monotone needs some kind of "partial pull" feature for this 
in order to work under all circumstances. It should be possible to just 
pull one single revision (e.g. always the latest) to do

 $ mtn pull mtn://my.host.net/h:my.branch

To avoid too complex URIs a selector without a prefix should default to 
a branch rather than "everything" what matches, so

 $ mtn pull mtn://my.host.net/my.branch

should be actually the same like

 $ mtn pull mtn://my.host.net/b:my.branch

which pulls all revisions for this branch.


Note that all this only addresses netsync communication, e.g. I do not 
particularily define a way how to access a certain file inside a certain 
revision (do we need this at all and where?).


I've put a wish for this already on the MtnSummit/Ideas page, feel free 
to edit / expand that. Other opinions?

Thomas.

--
ICQ: 85945241 | SIP: 1-747-027-0392 | http://www.thomaskeller.biz
> Guitone, a frontend for monotone: http://guitone.berlios.de
> Music lyrics and more: http://musicmademe.com