Re: [EGIT PATCH] Convert author and comment on demand.

On Sun, 3 Dec 2006, Jakub Narebski wrote:

This is why I _hate_ patches in attachments.

Sure, I can make my mail reader show the attachments. But when I "reply", I want the patch to be indented with the regular "> " thing and visible in the reply, so that I can say "I like the patch in general, but <this> part of it is seriously broken".

And yes, the personal mailreader I have has a "include attachements in reply" mode. But I don't generally want to include attachments in any reply, I just want it for _patches_.

[ Side ntoe: besides, many people send broken attachments that aren't 
  marked as text, but as 8-bit binary data or something, even if it's a 
  text-patch - so "attachement problems" are almost as common as the 
  non-attachement "line wrap" or whitespace problems are - people who 
  think that attachments automatically means that the thing is correct are 
  just deluding themselves.

  The fact is, you can get attachments wrong exactly the same way you get 
  linewrapping wrong. It's just that the pure binary data is likely to 
  always make it through correctly with an attachment, but if it gets 
  marked as a binary MIME-type, that doesn't much help, because while the 
  data is there, it's still going to act the wrong way for any 
  _discussion_ about it. ]

In other words: there are lots of things that make sense as true attachments. It's just that "patch" is not one of them.

Patches, unlike for example full files or tar-balls, by their very design are (a) text and (b) something where the whole _point_ is discussion and quoting about them. If we didn't want to discuss the patch contents, we wouldn't send them as patches in the first place, they'd be git-to-git transfers.

So this all boils down to one thing:

	PATCHES ARE NOT SEPARATE FILES TO BE ATTACHED TO AN EMAIL.

Patches are to be considered part of the discussion, not separate. And thus they should be in the main body of the email, exactly so that people (regardless of mailer settings and details like that) will automatically quote them, and they get passed around as integral to the discussion as the discussion itself.

I just jumped in because this is a pet peeve of mine. Some people seem to think that patches are "binary files" just because they have whitespace requirements and line-wrapping matters. But whitespace and line wrapping is important even in regular discussion, and patches really _are_ about the _human_ communication, just as much - and perhaps more so - as they are about feeding to the "patch" program.