Re: Waste of bandwidth? (Was: Let the header name be "Location:")

[Ned Freed <NED@xxxxxxxxxxxx>]

> >	Roy!  We're talking about RFC822 and MIME, here.  The byte-counting
> > war got lost 20 years ago.  Honest.

>   Just curious--is this a common view?  Also, do you eschew
>   byte-counting everywhere or do you differentiate between useful
>   and frivolous header fields?

Speaking for myself, I agree that the war was lost, but not all at once 20
years ago. Its more like it was lost in a series of little battles over the
entire period.

RFC821/RFC822 ecomonize on the amount of data vs. some alternative protocols
(e.g. the ones that carry the entire envelope around to all recipients), but
are wasteful in representing it in readable way. It now seems clear that making
things readable was a huge win, as evidenced by the fact that challengers
without this characteristic have completely failed to displace RFC821/RFC822.

MIME was built on top of the observation that attempts to do MIME-like things
in ways that aren't aligned with old-style RFC821/RFC822 usage had been tried
many times and had always failed. And history has repeated itself, in that thus
far MIME has won out against challengers, some of which did things in less
wasteful ways.

However, this doesn't mean that byte-counted protocols don't have a place in
the world. There are plenty of contexts where space and economy are major
issues (e.g. NFS), where it seems silly to consider a readable alternative.
(Note that I didn't cite ASN.1 or BER here...) Its all a question of picking
your battles carefully.

>   What about signatures?  PGP signatures are beginning to make the wasted
>   space by extra header fields pale in comparison.

This is part of the reason that the waste inherent in readable representations
has never been of major importance -- there are too many other things that are
far more wasteful. As far as PGP goes, at least it has some benefit. What about
the lovely little signature files so many people use? I got one the other day
that was a full two page graphic done in VT100 escape sequences! Cute, but is
it really worth it when sent to a mailing list?

>   With the explosive growth of the Internet, byte-counting in the
>   small will yield large rewards in the large.

I disagree. The trend now seems to be towards progressively larger messages,
and the trend is accelerating. I see large sites now whose traffic consists
almost entirely of word processor files and attached spreadsheets. Their
average message size is at least 32K if not 64K, and we're talking about high
traffic volumes as well. The cost of carrying this sort of stuff far exceeds
the cost of removing the few bytes that make your header more readable.

Put another way, while its possible that small efficiency improvements will
yield large rewards, its entirely dependent on where the improvements are made.
For example, as 8bit-capable SMTP comes into widespread use, one efficiency
improvement with a potentially massive yield is the definition and deployment
of a more efficient binary encoding than base64. Its possible to define an
encoding that's compatiable with 8bit-capable SMTP that introduces something on
the order of 3% overhead, as opposed to the 33% overhead of base64. The savings
to be had here dwarf those to be had by removing a few bytes from some header.
(There's a draft of this approach I expect to see posted any time now.)

				Ned

P.S. Don't hold your breath waiting for binary SMTP and the option of not
encoding anything. Its a defined service, but it has taken 5 years to create
8bit-capable agents, and binary-capable is a *much* harder problem.