Opinions on in-band variation signaling

[Aristotle Pagaltzis]

* Sean Leonard <dev+ietf at seantek.com> [2014-11-12 00:50]:
> At least one commentator (Larry Masinter) specifically requested that
> the variant identifier be included in-band in the Markdown content,
> rather than as part of the metadata. I disagree with in-band signaling
> (as has been registered earlier on this list).

Both have value.

    Ruby’s postulate:
    The accuracy of metadata is inversely proportional to the
    square of the distance between the data and the metadata.

That is to say, a value signalled in-band will be correct much more
often than one signalled via MIME/HTTP headers.

Also, in-band vs out-of-band simply address different scenarios.

An in-band signal travels with the data automatically, without any
special measures taken, whereas out-of-band metadata has a fair chance
of getting lost somewhere in transit. How is a user agent supposed to
store the variant parameter from the content-type after downloading
a Markdown file to disk?

But if you want to make decisions about the data during transit, having
to parse the payload to find the signal is prohibitively expensive. An
out-of-band signal is cheap, and is visible to a lot of infrastructure
that would never implement such specialised parsing of payloads such as
Markdown. (I mean, that is the exact same reason for having a MIME type
separate from text/plain in the first place.)

It’s nonsensical to portray this as a binary choice between two options.
Those are two separate choices and discussions.

Regards,
-- 
Aristotle Pagaltzis // <http://plasmasturm.org/>