Re: [ietf-types] Status of application/patch or text/patch?

[Simon Josefsson <[e-mail]>]

Julian Reschke <[e-mail]> writes:

> On 2012-07-19 14:12, Simon Josefsson wrote:
>> Jon Moore <[e-mail]> writes:
>>
>>> Hi folks,
>>>
>>> Given the recent arrival of HTTP PATCH in RFC 5789[1], it seems like
>>> registering a media type for the output of the 'diff' utility would be
>>> a good idea. I found this thread from 2007 where Julian Reschke
>>> proposed this exact thing:
>>>
>>> http://www.ietf.org/mail-archive/web/ietf-types/current/msg00591.html
>>>
>>> Does anyone know where this landed? I'd like to pick up the torch and
>>> at least work on text/diff (using the unified diff format, and limited
>>> to those diffs that can be rendered as valid text/* types). I've
>>> started putting together an Internet Draft for it, but was wondering
>>> if this had actually run into technical trouble, or if it just ran out
>>> of steam in 2007.
>>
>> The issue of how to deal with different character encodings were
>> unresolved, and I think it is important to get right.  The approach to
>> specify an application/patch is the easy way out.  Trying to resolve it
>> for a text/patch may be possible, but I fear it will be problematic in
>> practice.  Thus I lean towards a application/patch.
>
> Could you elaborate of what exactly you think would be easier to
> specify using application/patch instead of text/patch?

Some reasons that I had in mind were:

* There is no character set issues -- the encoder doesn't need to know
  which character encoding was used.  The specification doesn't have to
  talk about character sets (which makes up a big chunk of the current
  spec.)

* It can deal with diff's of files with different character sets,
  e.g. converting ISO-8859-1 files to UTF-8.  Like the draft says now,
  the text/* approach is unable to deal with this.

* Sometimes text/* parts are modified by gateways, and this is usually
  avoided by using an application/* type.  So the spec doesn't have to
  talk about these concerns.

However I support doing both media types.  For some situations,
text/patch is appropriate.

/Simon