[00:05:33] wseltzer joins the room
[00:06:49] wseltzer joins the room
[00:21:28] wseltzer leaves the room
[01:13:58] wseltzer leaves the room
[08:15:52] ilari.liusvaara joins the room
[10:06:08] roberto_peon joins the room
[10:21:39] rcross joins the room
[12:30:29] bortzmeyer joins the room
[12:43:27] rcross leaves the room
[12:51:27] rcross joins the room
[12:54:10] nakakura joins the room
[12:56:23] sho.y@xmpp.jp joins the room
[12:57:37] nakakura leaves the room
[12:59:46] nakakura joins the room
[13:01:06] ted.h joins the room
[13:01:48] Barry Leiba joins the room
[13:02:21] stpeter joins the room
[13:02:26]  audio is working fine.
[13:02:34]  roberto_peon: thanks for the report
[13:02:35] g.e.montenegro leaves the room
[13:03:29] mcmanus joins the room
[13:04:05] Julian joins the room
[13:04:53]  howdy
[13:05:36] frodek joins the room
[13:05:53] uǝsuɐɥ ʎuoʇ joins the room
[13:09:16]  this is http://www.ietf.org/proceedings/89/slides/slides-89-httpbis-5.pdf
[13:09:49]  Put "mic:" if you want me to channel you in the room.
[13:10:12]  Thanks for relaying Barry.
[13:10:31] Mark Donnelly joins the room
[13:11:04] loreto.salvatore joins the room
[13:11:12] michael.r.sweet joins the room
[13:11:44]  mic: pausing is served via flow control right now, and this flows through all intermediaries, unlike priorities.
[13:12:59] SM joins the room
[13:14:57]  http://www.ietf.org/proceedings/89/slides/slides-89-httpbis-4.pdf
[13:15:16] Dan Wing joins the room
[13:15:26] Adam Bishop joins the room
[13:15:54] Julian notes that Firefox has really trouble with these PDFs
[13:16:13] roberto_peon notes it works OK on Chrome.
[13:17:04] g.e.montenegro joins the room
[13:17:20] npdoty joins the room
[13:17:21]  not on f20! (but I am running unstable)
[13:17:26] wseltzer joins the room
[13:18:54] barry.constantine2 joins the room
[13:19:20]  hello, just checking jabber
[13:19:33]  jabber is working.
[13:19:34] rcross leaves the room
[13:21:19] barry.constantine2 leaves the room
[13:21:36]  yes.
[13:21:45]  that is still where we're at.
[13:21:56]  You want that relayed to the mic, Roberto?
[13:22:12]  only if necessary. If I feel it *must* be relayed I'll prepend "mic:"
[13:22:13]  :)
[13:23:34]  Larry Masinter at the mic
[13:24:09] barry.constantine2 joins the room
[13:24:22] Steve Olshansky joins the room
[13:25:33] barry.constantine2 leaves the room: I'm happy Miranda IM user. Get it at http://miranda-im.org/.
[13:25:58] dougt joins the room
[13:26:06]  speak up plz. :)
[13:27:21] barry.constantine2 joins the room
[13:27:23] Jxck joins the room
[13:27:25]  mic: Hasan and Will have access to a set of requirements that I sent out (but apparently not to the list) that might be interesting to discuss.
[13:28:28] barry.constantine2 leaves the room: I'm happy Miranda IM user. Get it at http://miranda-im.org/.
[13:28:44]  Is it really "strictly"?  I thought the whole nature of the proposal was advisory.
[13:30:18] uǝsuɐɥ ʎuoʇ leaves the room
[13:31:08]  because it isn't rasntitive across the proxy, yea.
[13:31:12]  *transitive
[13:34:36]  mic: Not during the timeframe.
[13:36:08]  mic: To reinforce, we know what we have right now is not adequate.
[13:36:21]  It is pretty terrible across proxies.
[13:36:41]  "Would you rather have nothing or the other guys' proposal" is often a useful question at moments like these.
[13:38:22]  no
[13:38:30] Adam Bishop leaves the room
[13:39:57]  DAG (in the worst case).
[13:40:19] m&m joins the room
[13:40:48]  The priority scheme with 'flat' levels in groups does not traverse proxies well. It becomes very very difficult to make that scheme make sense across proxies.
[13:41:35]  mic: (above) + what Will is saying is far far worse at proxies, where you're forced to discretize and allocate portions of the space to different clients, but that is difficult since you don't know how many clients you'll have, and thus what the partitioning is..
[13:41:46]  (I'm agreeing with Will)
[13:42:31]  mic: The browser case is easy compared to the proxy case.
[13:42:39] stpeter leaves the room
[13:43:33]  ... like me? :)
[13:45:05] richgold joins the room
[13:46:18] stpeter joins the room
[13:46:55]  Question was:  do you like one of these or dislike on of them or the other.
[13:47:07] rigo joins the room
[13:47:11]  mic: This isn't a small deal. I'd be happy to claim publicallty that HTTP/2 is broken if we spec something that is so easily broken/gamed.
[13:47:17]  yes.
[13:47:59]  NO
[13:48:26]  ?me waves
[13:48:30]  Anyone out there relaying?
[13:49:03]  ()*&!#&^(*)&(*)98-0
[13:49:35]  Sorry, I was at the mic line, even if I didn't end up speaking.
[13:49:41]  CAn you break in.
[13:50:18]  I was trying to get that expressed for the better part of 10 minutes.
[13:50:46]  yes, happy with the current pull request
[13:51:04]  Sorry, we're also apparently lagged.
[13:51:09]  :)
[13:51:15]  Dangers of remote participation :)
[13:51:28]  Sorry, Roberto: I was working on something else, and stopped paying attention to the jabber room.
[13:51:35]  Mea culpa; mea maxima culpa.
[13:51:37]  it happens :)
[13:51:54]  I'm back now.
[13:52:03]  Barry Leiba: I can relay if you'd like
[13:52:18]  PSA: If I don't jump up, yes, please back me up.
[13:52:25]  http://www.ietf.org/proceedings/89/slides/slides-89-httpbis-7.pdf
[13:52:31]  k
[13:53:28] bartosz.balazinski joins the room
[13:56:45] Steve Olshansky leaves the room
[13:57:54]  mic: ekr: Did you agree that, if the max parallelism (i.e. MAX_CONCURRENT_STREAMS) limited is sufficient to add back observability to the server within whatever constant bounds? Also: Agreed that one cannot effectively make a blacklist.
[13:58:07]  *is limited
[13:58:34]  hehehehe.
[13:58:35]  Barry is getting up, Roberto
[13:58:39]  thx
[13:59:17]  martin could speak closer to the mic
[14:00:31]  Genearting the new page requires TLS authorization :)
[14:00:35] Eliot Lear joins the room
[14:02:41] Steve Olshansky joins the room
[14:03:04]  Great point.
[14:03:35]  Yup. it is worthless.
[14:04:16]  mic: Careful it isn't perfect unless groups (or referrer) are sent along with the request, thanks to intermediaries.
[14:04:38] dougt joins the room
[14:04:45] dougt leaves the room
[14:04:48]  Barry is getting up, Roberto
[14:04:51]  thx
[14:06:14] loreto.salvatore leaves the room
[14:06:51]  "There are intermediaries which muck around with" could be followed will with *, as far as I can tell.
[14:07:06] Eliot Lear leaves the room
[14:07:09] Eliot Lear joins the room
[14:07:48]  Yes, the first variant is strictly more easy to understand :)
[14:08:27]  don't you just hide it from the very next guess?
[14:09:48] dougt joins the room
[14:09:59] dougt leaves the room
[14:10:33] dougt leaves the room
[14:11:03]  the text on the screen doesn't really makes sense. I doubt it is readable beyond the first 2 rows 	  				 	 	 	    		  	   		  		
[14:11:29] dougt joins the room
[14:11:31] m&m leaves the room: Disconnected: closed
[14:11:32]  has someone the URI of that email? 	  				 	 	 	    		  	   		  		
[14:11:48]  mic: The most important thing right now is signaling that is durable across intermediaries about the incompressibilty of something.
[14:11:55]  This requires a single bit in the headers.
[14:12:14]  and by signaling I mean signaling that something should not be compressed.
[14:12:32]  in q
[14:12:36]  thx Barry
[14:14:18] m&m joins the room
[14:15:09]  http://lists.w3.org/Archives/Public/ietf-http-wg/2014JanMar/0838.html
[14:15:16]  mic: A single bit in the headers is more efficient. You don't want to compress *any* of that data for a CORS request against one that isn't.
[14:15:25]  yes.
[14:16:05] Eran Messeri joins the room
[14:16:10]  ekr's comment.
[14:16:24]  I guessed right.
[14:16:28]  :)
[14:16:52]  We don't need to redo the opcodes.
[14:17:36]  micL: yes please.
[14:19:21]  which issue?
[14:19:34]  363
[14:19:38]  thx
[14:20:23]  mic: loveit. HATE renegotiation.
[14:24:05]  GOAWAY with some status code, probably NOT HTTP-code, since it isn't HTTP specific?
[14:24:10]  on the lsit sounds good
[14:24:21] bizzbyster joins the room
[14:24:33] bizzbyster leaves the room
[14:26:58] Peter Lepeska joins the room
[14:27:23]  TRON'
[14:32:36]  mic: We have experience with alt-protocol.. which works, and is in use today.
[14:33:48] Peter Lepeska leaves the room
[14:34:05]  ekr has covered my comment.
[14:34:12]  Okay, thanks
[14:34:49] Steve Olshansky leaves the room
[14:35:01]  alt-svc :)
[14:35:10]  (not alt-proto!)
[14:35:50] Peter Lepeska joins the room
[14:37:37] ted.h leaves the room
[14:38:02]  mic: and for CDNs..
[14:38:15]  (good stuff, full support here, yes put in the spec)
[14:40:23] Steve Olshansky joins the room
[14:41:31]  mic: Yes. This is something within our direct experience, and is necessary for the long-term supoort/stability of servers under HTTP2
[14:42:05] loreto.salvatore joins the room
[14:42:47] ted.h joins the room
[14:43:36]  The server would close the connection in our case if the client didn't implement it. If the client didn't follow it enough, it might be considered an attack, which would suck.
[14:44:09]  mic: and/or it might not be an HTTP request.
[14:44:18]  We should also be careful about using Alt-Svc as a replacement for a 3xx status code
[14:44:27]  e.g. there may be no outstanding HTTP request when we need to do this.
[14:44:52] ted.h leaves the room
[14:45:01] ted.h joins the room
[14:45:15]  long mic q.  Shall I wait, or break in?
[14:45:22]  I'd imagine that the way this was implemented was send the recommendation to COME_BACK_ELSEWHERE followed by a GOAWAY.
[14:45:34]  If you think it is interesting, yes, but sounds like no from here :)
[14:45:40]  right
[14:45:41]  (re: breaking in line)
[14:46:09] m&m leaves the room
[14:47:32]  That isn't necessary for this since one can simply wait some time for the new connection to be established and then get the current connection torn down.
[14:47:44]  (by this I mean information about the TLS state)
[14:49:08]  To reiterate, we need this as a frame in HTTP2 because there is no guarantee that there is an outstanding *http* request when one wants this to happen
[14:49:11] Julian notes that the example on the screen is syntactically incorrect
[14:52:01]  mic: But that takes another RTT
[14:52:04]  :)
[14:52:52]  rtt=return to table
[14:53:00]  heh
[14:53:21]  Important points only for relay at this point.
[14:53:27]  7 minutes left in the session.
[14:54:09] Eran Messeri leaves the room
[14:55:23]  both GOWAWAY and this are compelling.
[14:55:40]  for http/1.1
[14:56:16] rigo leaves the room
[14:56:25] Peter Lepeska leaves the room
[14:56:54]  In case I don't get a chance to say before the endo fo the session: Barry, Ted, thanks for relaying!
[14:57:19] carlos.paparoni joins the room
[14:59:27]  You're welcome, always,
[14:59:37] Barry Leiba leaves the room
[14:59:59] ted.h leaves the room
[15:00:00] Eliot Lear leaves the room
[15:00:04] Julian leaves the room: Computer went to sleep
[15:00:17] Jxck leaves the room
[15:00:24] SM leaves the room
[15:00:30] carlos.paparoni leaves the room
[15:00:32] loreto.salvatore leaves the room
[15:00:39] Steve Olshansky leaves the room
[15:00:46] frodek leaves the room
[15:00:56] nakakura leaves the room
[15:00:56] roberto_peon leaves the room: offline
[15:01:17] dougt leaves the room
[15:01:17] npdoty leaves the room
[15:01:20] sho.y@xmpp.jp leaves the room
[15:01:24] michael.r.sweet leaves the room
[15:01:49] mcmanus leaves the room