[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [RRG] Why delaying initial packets matters



Eliot,

On Feb 11, 2008, at 1:55 AM, Eliot Lear wrote:
An obvious first candidate would be DNS I think.

The DNS provides a useful model for a pull-based system, but I'm not sure it would be particularly useful to assume if something can be shown it can't work if you put the mapping into the DNS, that implies it can't work period.

Keeping in mind that these queries are likely recursive, we shift our scrutiny to the name server. Supposing a query is dropped, will the next query go to the same name server or to another listed name server? If it's to another name server could that query also be dropped? Will behavior differ based on the number of zone cuts and whether the appropriate name servers are in the same proximity as the parent servers?

It's amazing that the DNS works at all, isn't it? And yet, with all that recursion and potential for timeout, it seems to work reasonably well.

However, as I've said on several occasions, while I might personally use the DNS as a model for a pull-based system (primarily because most folks think they understand what that implies), I don't think the DNS (as it exists today) is the right answer. For one thing, the timeouts would likely need to be rejiggered since the existing timeouts are based on assumptions that were derived from when 56 Kbps was considered "high speed".

As far as I can tell, a pure pull-based system is likely as unworkable as a pure push-based system. As Robin has suggested, a hybrid system is probably going to be necessary, so there are yet more tradeoffs to consider.

Regards,
-drc


--
to unsubscribe send a message to rrg-request@psg.com with the
word 'unsubscribe' in a single line as the message text body.
archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg