People intend to move in the opposite direction, see #27809 (comment)

Oh no

I mean I could just reverse the regexp…

I wouldn't hurry with such mass replacements. I personally can't see why it should make a significant difference if you use one way or another one.

@vcunat The main reason in my opinion is consistency. Having all packages follow the same pattern is a lot less confusing for new users. Of course there is an almost infinite amount of ways you can package software, but IMO we should have some common ground for some parts. Maybe this PR isn't the proper location to discuss this and this is probably not the best time as well, as a new release is pretty close and there is more important stuff to do.
But these inconsistencies and "we'll refactor that one day"-things tend to add up and I always try to fix/refactor/… them ASAP.

Perhaps, but first we need a clear consensus that one of the two variants is preferred ;-)

Yes, I agree with you. What do you think is the proper place for such a discussion? Discourse? IRC? Private Mails? This PR? A RFC?

This thread isn't a bad start, IMHO. RFC is another option. We could link to that thread from other places if it seems there are too little eyes.

Consistency: note that we have URLs inside fetchurl and similar calls very commonly, and there we often want to interpolate some variables, and that can't be done in the unquoted style ATM.

Yes, I also found that the url type fails for the characters I mentioned above: $#;(. While $ is mostly used when using variables in a URL (which is a legitimate thing to do of course), #, ;, and ( are used as well. So these URLs would have to be strings (or am I just missing some escaping rules?) meaning that using the URL type will always include inconsistencies.

I haven't found other places where I can link this thread yet, but I'll keep looking :)

And we do need quoted URLs for src handling (putting ${version} inside the URL string is a good idea from the point of view of avoiding sily mistakes), so full unification means either quoted URLs or rejecting useful tools.

While I'm favor of not quoting URLs, this PR causes a lot of code churn for a minor syntactic cleanup, which I think is to be avoided.

(BTW, I don't agree with the sentiment that "People [who?] intend to move in the opposite direction".)

@edolstra

People intend to move in the opposite direction, see #27809 (comment)
People [who?] intend to move in the opposite direction

Most people who commented in #27809, duh.

Closing this as most seem to be against this change (me included). Maybe eventually this could be enforced with some ofborg check to disallow unquoted URLs.