Re: address@hidden: 23.0.0; (thing-at-point 'url) returns invalid urls]

[Stefan Monnier]

>> >> Would someone please DTRT and ack?
>> > There's no problem here and nothing to do. The URLs returned are
>> > perfectly valid. See the discussion that followed the original email.
>> 
>> Maybe the docstring (and or behavior) should be adjusted to make the
>> distinction clear between "assuming there's a URL at point, return it" and
>> "check if there's a URL at point, and if there is, return it".
>> 
>> This is often a somewhat subtle but important difference.  When the user
>> does M-x browse-url RET, it makes perfect sense to place as default in the
>> minibuffer the best URL we could come up with, even if it doesn't
>> look very
>> likely to a good one.
>> 
>> OTOH if we create a command which can do two different thing depending on
>> whether there's a URL at point, we would want to be much more stringent on
>> what we consider as an acceptable "URL at point".

> Seconded.

> The doc can distinguish between the URL itself being valid (syntactically)
> and the URL having a live target. The first is a property of the URL; the
> second is a property of its target.

No, I think this would be a mistake.  We're still talking only about URLs
independently from their target.  Liveness of a URL target is a property
that is difficult/impossible to ascertain and can change at any moment.
I don't want `browse-url' to ignore "http://www.iro.umontreal.ca"; under
point just because I'm currently not connected to the internet or because it
is temporarily down, and neither do I want it to wait for a timeout before
deciding it.

The difference we want to talk about instead is between something that
heuristically is likely to be meant as a URL and something which may be used
as a URL but doesn't particularly "look like a URL".


        Stefan