Re: [hybi] frame length encoding

["Shelby Moore" <[e-mail]>]

> On Sat, 21 Aug 2010, gustav wrote:
>
>>> I think I would prefer a plain 32 bit size, with no special treatment
>>> for
>>> smaller fragments.
>
>> Do you care to elaborate why ?.
>
> I think most of the arguments around the sizes already have been mentioned
> and
> I'm not sitting on any new secrets.
>
> I think sizes larger than 32bit will be very rare for Websockets, even for
> a
> long time to come into the future. Thus a vast vast majority of all frames
> will be sent with at least 4 bytes set to zero (with a fair amount of them
> not
> even implementing >32 bit support). In my mind, I think the few users that
> acually want to send more than 32bits of data can send it in multiple
> packets
> to get around the limitation. It's not like having a 32bit limit here will
> prevent anyone from sending more than 4GB of data.


No because for any file on disk larger than 4GB, we will have to move the
entire file through the CPU, when we wouldn't had to otherwise:

http://www.ietf.org/mail-archive/web/hybi/current/msg03402.html

More abstract explanation of the above link:

http://www.ietf.org/mail-archive/web/hybi/current/msg03410.html
http://www.ietf.org/mail-archive/web/hybi/current/msg03417.html


> I don't think 4 bytes size is any particular size "loss" for small packets
> to
> feel bad about, so I prefer the simplicity of always just getting the size
> in
> a single specified way.
>
>> Its not the lazy coders who pay others net bills.
>
> Uh?

Imagine a WebSocket used to push wakeup events to real-time application
(game perhaps), so that is only perhaps 1 byte of payload. You are
proposing to double the network traffic.