Zotonic
Zotonic
zotonic@conference.zotonic.com
Thursday, 11 October 2012< ^ >
arjan has set the subject to: Zotonic - the Erlang Content Management Framework
Room Configuration

GMT+2
[00:04:14] arjan leaves the room
[00:05:33] arjan joins the room
[01:11:56] arjan leaves the room
[01:12:36] arjan joins the room
[01:20:06] arjan leaves the room
[01:20:15] arjan joins the room
[01:30:05] arjan leaves the room
[01:33:28] arjan joins the room
[02:27:05] arjan leaves the room
[02:27:44] arjan joins the room
[03:26:30] arjan leaves the room
[03:29:30] arjan joins the room
[05:44:26] arjan leaves the room
[05:45:59] arjan joins the room
[09:06:11] andreas.stenius joins the room
[09:17:11] andreas.stenius leaves the room
[09:43:03] Ilya Rezvov leaves the room
[10:36:06] Ilya Rezvov joins the room
[10:50:44] maas.maarten.zeeman joins the room
[10:57:42] <maas.maarten.zeeman> hunting down strange httpc tls bug. seeing packet loss on connections and unknown records when inspecting with wireshark. something is wrong.
[11:02:01] <Marc Worrell> hmmm, I also got some TLS errors when sending e-mail
[11:02:38] <Marc Worrell> When I remember correctly it was something with the certificate. But I ignored it as it was just a mailing list and only a few recipients.
[11:03:17] <maas.maarten.zeeman> Probably due to the retransmissions I get multiple responses to 1 request.
[11:04:03] <maas.maarten.zeeman> checking things with curl now
[11:05:07] <Marc Worrell> hmmm
[11:05:14] <maas.maarten.zeeman> that uses openssl too
[11:05:31] <Marc Worrell> you don't use new SSL ?
[11:06:20] <maas.maarten.zeeman> standard httpc. ill have to check what's inside
[11:06:56] <maas.maarten.zeeman> I also get unknown messages being sent to the httpc_manager {ref-soemthing, ok}
[11:10:46] <maas.maarten.zeeman> curl handles it just fine.
[11:11:50] <Marc Worrell> Might be a new_ssl thing
[11:12:10] <Marc Worrell> R14B uses new ssl instead of the old openssl based ssl
[11:14:28] <maas.maarten.zeeman> aha. not nice.
[11:14:39] <Marc Worrell> scales better :p
[11:15:27] <maas.maarten.zeeman> sure it does :-) openssl can do hardware crypto (if you buy the right hw)
[11:15:48] <Marc Worrell> :p
[11:15:59] <Marc Worrell> add a small 'if' in the code :p
[11:16:05] <maas.maarten.zeeman> pff. that must be it
[11:16:33] <Marc Worrell> maybe check if there are ssl fixes in the newer releases?
[11:16:49] <Marc Worrell> or see how you can switch back to old ssl (fow now)
[11:17:00] <maas.maarten.zeeman> with the old ssl it works. sigh
[11:17:42] <maas.maarten.zeeman> probably some error handlers masking errors
[11:18:08] <maas.maarten.zeeman> the repeated responses also point at a state not updated or something
[11:18:27] <maas.maarten.zeeman> thanks
[11:21:54] <maas.maarten.zeeman> R15
[11:23:14] <Marc Worrell> ah, you could switch back>
[11:23:17] <Marc Worrell> ?
[11:24:11] <maas.maarten.zeeman> I think so. trying that. It is possible to switch between with the cipher_suites call
[11:24:55] <maas.maarten.zeeman> cipher_suites(erlang | openssl)
[11:25:26] <maas.maarten.zeeman> oh, this is R15 documentation haha
[11:27:59] <maas.maarten.zeeman> hmm, that doesn't switch to old ssl.
[11:30:01] <Marc Worrell> but it might use openssl, which should be good enough?
[11:31:52] <maas.maarten.zeeman> dunno, but i think i am warm... I can try it on the console and see what wireshark returns.
[11:33:57] <maas.maarten.zeeman> with the new it is erlang managing the ssl socket and not openssl?
[11:34:14] <Marc Worrell> yep
[11:34:33] <Marc Worrell> at least doing all the cipher and protocol work
[11:34:50] <maas.maarten.zeeman> aha, and then they pass the cipher stuff to openssl
[11:36:06] <maas.maarten.zeeman> that is probably where the unknown message is coming from then
[11:46:31] <maas.maarten.zeeman> the server i'm getting the errors on also does a redirect with a relative pathname. httpc is not buying that.
[12:11:00] <Marc Worrell> that is a quite bad server
[12:11:36] <Marc Worrell> saw that too on flickr, when there was an error redirect
[12:23:52] andreas.stenius joins the room
[12:52:25] arjan leaves the room
[13:35:40] <maas.maarten.zeeman> It is happening all over the place, indeed for error redirects. funda.nl has the same.
[13:37:12] andreas.stenius leaves the room
[13:37:14] andreas.stenius joins the room
[13:37:15] <maas.maarten.zeeman> this was a logon redirect
[13:37:25] andreas.stenius leaves the room
[14:08:19] maas.maarten.zeeman leaves the room
[14:42:20] maas.maarten.zeeman joins the room
[14:42:43] arjan joins the room
[14:43:14] <maas.maarten.zeeman> nice httpc masks ssl implementation switch options
[15:21:48] andreas.stenius joins the room
[15:39:34] arjan leaves the room
[15:42:01] arjan joins the room
[15:42:45] arjan leaves the room
[15:44:05] arjan joins the room
[15:49:07] arjan leaves the room
[15:59:53] arjan joins the room
[16:01:38] <maas.maarten.zeeman> the bug is in another location
[16:01:48] <maas.maarten.zeeman> old ssl and new ssl doesn't make a difference
[16:23:55] arjan leaves the room
[16:54:54] <maas.maarten.zeeman> found it. i'm doing asynchronous http requests... but if you do a httpc request in handler function (to handle a redirect) things go wrong with ssl. You can get two responses back. Totatlly weird. Just not call httpc from the handler then.
[17:20:09] maas.maarten.zeeman leaves the room
[17:52:42] Ilya Rezvov leaves the room: Replaced by new connection
[17:52:42] Ilya Rezvov joins the room
[18:10:29] arjan joins the room
[18:11:20] Ilya Rezvov leaves the room
[19:39:22] arjan leaves the room
[19:39:38] arjan joins the room
[22:01:18] Maas joins the room
[22:06:16] andreas.stenius leaves the room
[22:16:47] Maas leaves the room
[22:17:17] maas.maarten.zeeman joins the room
[22:28:15] arjan leaves the room
[22:30:35] arjan joins the room
[22:32:44] Ilya Rezvov joins the room
[22:37:45] Ilya Rezvov leaves the room
Powered by ejabberd Powered by Erlang Valid XHTML 1.0 Transitional Valid CSS!