Re: [hybi] Frame size

[Mike Belshe <[e-mail]>]

On Sun, Apr 18, 2010 at 5:23 PM, Ian Hickson <[e-mail]> wrote:

> On Sun, 18 Apr 2010, Mike Belshe wrote:
> > >
> > > N years from now (for some value of N) people will be saying the same
> > > about 32 bits. That's why I used a packed integer instead of a
> > > fixed-width integer to specify lengths.
> >
> > Of course, agree.  And despite where 32bits has failed us in history
> > previously, I still think this is far enough out that we'll be able to
> > upgrade by then.
>
> Sounds like Y2K reasoning. :-)
>

It's not at all the same.  As you stipulated below, we're talking about the
edge case of super large messages (greater than 4GB!).  Even in the future,
it will be an edge case, and it has an easy workaround (multiple frames).
 Y2K was the common case; once N years elapsed, everyone had the problem in
the common case.  Lastly, it should be noted that everyone *did* upgrade for
Y2K and it wasn't the end of the world.  A better example might be the IPv4
to IPv6 migration...  It should have been the end of the world, except, it
wasn't. :-)

Fixed lengths has a lot of benefits - including for the 'ease of
programming'.



>
> > If you think N is sufficiently small for 32bits, we should use 64bits,
> > would that make N sufficiently far away? :-)
>
> Reserving more than four bytes every time just seems like a lot of wasted
> bytes. Not that they're expensive or anything, but most of these frames
> are going to be only a few bytes, so it just seems silly to be padding
> everything full of zeros.
>

I wasn't proposing it seriously.

Mike


Mike


> --
> Ian Hickson               U+1047E                )\._.,--....,'``.    fL
> http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
> Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'
>