[dns-operations] Configurable TC=1?

Edward Lewis edward.lewis at icann.org
Mon Dec 21 18:57:17 UTC 2015


On 12/21/15, 13:13, "Robert Edmonds" <edmonds at mycre.ws> wrote:
>Not to beat a long gone horse, but if you need to make a large number of
>requests, using blocking I/O and a thread-per-connection model for a
>connectionless protocol is a really bad idea.  Why didn't you use a
>single UDP socket to send queries and process responses as they came in?

I did.[0]

No matter how you slice it, where any probing code will sit is waiting for
non-responded queries.

Simply because it doesn't sit for the responded queries. ;)

[0] - the code was written a long time ago, if I recall correctly I didn't
use any modern thread library.  I did encode into the 16bit ID the
necessary pointer to the query so when the response came back I'd know
what, who and why I'd asked.  Still, I ran out of space waiting on the
non-responses.  In staring at the code back then, I came to realize that,
if you watch the 'tail -f' of any network inspecting code, the log message
you'll always see is "waiting on response from...".  If there was no
waiting, the screen would scream by so fast you'd see nothing.

Kinda zen like.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4604 bytes
Desc: not available
URL: <http://lists.dns-oarc.net/pipermail/dns-operations/attachments/20151221/c828ac8e/attachment.bin>


More information about the dns-operations mailing list