[RFC] [PATCH] rewrite host.name:path as ssh://host.name/path

[Peter Arrenbrecht]

On Tue, Mar 18, 2008 at 10:14 AM, Alexis S. L. Carvalho
<alexis at cecm.usp.br> wrote:
> Thus spake Peter Arrenbrecht:
>
> > On Tue, Mar 18, 2008 at 8:40 AM, Alexis S. L. Carvalho
>  > <alexis at cecm.usp.br> wrote:
>  > > People often ask why "hg push user at example.com:path/to/repo"
>  > >  doesn't work.  Here's a patch that makes it work.
>  > >
>  > >  Basic heuristic:  if we couldn't find the scheme (http/file/ssh/...),
>  > >  there's a colon after the second character (to avoid problems with drive
>  > >  letters on windows), there's no os.sep before the colon and the path
>  > >  doesn't exist, assume it's a "hostname:path" ssh repo and translate it
>  > >  into ssh://hostname/path .
>  > >
>  > >  To specify the port, you'll have to use the current URL syntax.
>  >
>  > Hmm. Wouldn't it be better in the long run if instead you encouraged
>  > people to use proper URLs? This will become part of the public API.
>
>  The goal of the patch is to make things easier for users.  The
>  "hostname:path" syntax is much more common than ssh://... and people
>  still trip over the double slash thingie (and I'm not even sure that the
>  RFC(?) that recommended it is still valid).
>
>  And I have some vague memories that an "ssh://" URL was supposed to mean
>  "open a shell in that machine" without specifying what should happen
>  when there's a path.
>
>  So it's not like ssh:// is a standard.

I'm not saying it's a standard. I'm saying it's defined in the context
of hg without any guesswork.

>  >                                                                     So
>  > your detection logic could be used to guess they meant ssh:// and give
>  > them a hint. Could also give them a hint for .hgrc [paths] at the same
>  > time in case they're bothered by the long URLs.
>
>  I always find it "funny" when a program tells me "Hi! You said 'foo',
>  but I'm pretty sure you meant 'bar'. If you really wanted 'bar', say
>  'bar', OK?" and then fails with some kind of "invalid use of 'foo'"
>  error.  I.e. it's confident enough that I meant 'bar' to go through the
>  trouble of showing me that message, but not enough to actually use
>  'bar'.  That's fine when there could be catastrophic consequences, but I
>  don't think that's the case here.

Sure, that bugs me too. But I really think guessing is ok when facing
a user, but less desirable when it ends up in scripts and whatnot,
meaning it's going to be very hard to change. Can we distinguish this
here? I doubt it.

Anyway, just my 2 cents.
-peo