MIME type for diffs, *any* MIME type

[Julian Reschke]

Lisa Dusseault wrote:
> Hi,
> 
> This has come up before and stalled the whole attempt to standardize a 
> PATCH method for HTTP: the lack of a properly registered MIME type for 
> any binary or text diff format.
> 
> The GDiff format described in http://www.w3.org/TR/NOTE-gdiff-19970901 
> states the MIME type but it was never registered.  The IPR situation is 
> completely unclear.  No author of this has ever responded to my pings.
> 
> The VCDIFF format is described in http://www.faqs.org/rfcs/rfc3284.html 
> does not choose a MIME type. The IPR disclosure 
> <https://datatracker.ietf.org/public/ipr_detail_show.cgi?ipr_id=40> 
> states that AT&T grants licenses to "transmit data over the Internet 
> under HTTP 1.1".   I do not want to start to interpret whether that 
> covers new HTTP 1.1 methods as well as the ones that existed when this 
> grant was made.  An AT&T guy responded to my emails and said he'd 
> suggest a more liberal license to the lawyers but we never saw anything 
> come of that.
> 
> Julian recently mentioned the POSIX 'diff' format which is most widely 
> used and probably under the most liberal license due to existing source 
> code licenses.  No official MIME type was ever suggested for this until 
> Julian's mail but a bit of googling suggests that some applications 
> informally use application/diff and this is probably what is meant by 
> that.  I am unaware of an existing formal spec for this -- and that's 
> the sticking point for registering this one -- although clearly 
> interoperability is common.
> 
> Can we just ask IANA to register any or all of these, perhaps with 
> appropriate disclaimers in the form about IPR or specification 
> formality?  Surely that would be better than the status quo.  Can I help 
> cut red tape here?

I would propose the following:

- Write an RFC registering a media type (or media types) for the various 
output formats of POSIX diff (see previous thread over here).

- Define a *very* simply binary patch format, without looking at GDIFF 
or VCDIFF. I'm ready to work on that.

Best regards, Julian