Re: [tcpm] WGLC: draft-ietf-tcpm-urgent-data-02.txt

[David Borman <[e-mail]>]

Fernando and Wes,

On Feb 11, 2010, at 9:53 PM, Fernando Gont wrote:

> Hello, Wes,
> 
> Thanks so much for your feedback. COmments in-line....
...
>> Semi-Technical:
>> 
>> - section 2.3 - on the allowed length of urgent data, should this
>>  mention that obviously there is a limitation due to the length
>>  of a sequence number field?  You can't send a terabyte of urgent
>>  data.
> 
> Well, you cannot send a terabyte of normal data, either. :-)
> 
> I see your point, though... Here we just mentioned what was there in RFC
> 1122...
> 
> I think the spirit of the text in RFC 1122 is that the mechanism (flag +
> pointer) should not limit the amount of urgent data that you can send...
> (although, as dicussed on the list some time ago, the concept of "urgent
> data" is flawed. It's more of a "mark of an interesting point in the
> data stream")

Ah, you can do both over a TCP connection: send a terabyte of normal data, and send a terabyte of urgent data.  The urgent date is all the data before the urgent pointer.  So, just as TCP cycles through the sequence space to send a terabyte of normal data, so too it would cycle through to send everything in urgent mode.  Since there is only one urgent pointer that gets updated with each new packet, as long as the urgent pointer points beyond the end of the data in any given packet, the TCP connection will remain in urgent mode until it gets to the end of the terabyte of data. :-)  This is the technique documented in RFC 2675 "IPv6 Jumbograms", section 5.2.

So, there is no limit on how much data you can send while in urgent mode, from either the application or TCP point of view.

			-David Borman