> Could the user program specify in its query to the resolver, that it
wants to specifically use TCP?
Depends. getaddrinfo doesn't, but the DnsQuery_* APIs on Windows [1] and
res_info() on Linux [2] and Apple platforms [3] do. It doesn't look like
Android does [4] in its own resolver.
Thanks,
Tommy
[1]
https://learn.microsoft.com/en-us/windows/win32/dns/dns-constants#dns-query-options
[2] https://linux.die.net/man/3/res_query
[3]
https://developer.apple.com/library/archive/documentation/System/Conceptual/ManPages_iPhoneOS/man3/res_query.3.html
[2] https://developer.android.com/reference/android/net/DnsResolver
On 7/6/25 14:11, Michael De Roover wrote:
Hi,
When I looked at this I-D, the first thing that came to my mind is
performance differences between UDP and TCP (I’d imagine UDP to be
faster, but admittedly moot if it needs to be redone over TCP anyway).
The second thing was that at least every server I’ve ever dealt with
can already talk over both TCP and UDP. Given that, wouldn’t it make
more sense to put the onus on the DNS client (the resolver in 1035’s
diagrams I think) to start with TCP if it needs to?
Could the user program specify in its query to the resolver, that it
wants to specifically use TCP?
Met vriendelijke groet,
Michael De Roover
Mail: [email protected]
Web: michael.de.roover.eu.org
On 6 Jul 2025, at 20:37, Paul Wouters <[email protected]> wrote:
“ These complications can be avoided by assuming Classic DNS over TCP
is the only form of Classic DNS that new protocols need to accommodate.”
This is not how protocols using DNS work. You can’t say “new”
protocols must use only a specific flavour of DNS transport as
it’s mostly not up to the new protocol or application how DNS is
resolved.
This is at least 10 years too soon.
Paul
_______________________________________________
DNSOP mailing list -- [email protected]
To unsubscribe send an email to [email protected]
