Re: [dnsext] draft-mohan-dns-query-xml-00.txt

[Paul Vixie <[e-mail]>]

On Monday, October 03, 2011 17:34:48 Wessels, Duane wrote:
> Robert's idea was to take a (binary) UDP message, express it in hex, and it
> becomes the URL-pathname.  I don't see why you can't also do that with
> UPDATE.

theoretically we could, yes.

> (I'd add a message TCP-like length prefix, as you suggested, so the
> receiver knows it got the whole thing).
> 
> Maybe your point is that URL length becomes a problem?  We've all seen
> very long URLs I'm sure.

i'm not sure i've seen 64KB URL's.  i have seen 64KB UPDATE's.  but it's not 
the size that concerns me, rather the content-aware routers out there who may 
or may not respect our http cacheability headers.  we have no need of another 
layer of caching at the transport layer, especially if it means we have to 
translate our DNS TTL's into http expiration times.  and i don't know how to 
reliably avoid that caching on this data path if we use GET.

> Anyway, my preference for GET over POST is not that strong.

does anyone here have anecdotal evidence of POST being intercepted by a 
content aware router, or otherwise painfully middleboxed the way GET so often 
is?  at the moment i've got three weak reasons for preferring POST, but if any 
of the three is silly, they would not add up to a strong preference as they do 
(for me) at the moment.

paul
_______________________________________________
dnsext mailing list
[e-mail]
https://www.ietf.org/mailman/listinfo/dnsext