Re: [hybi] Last Call: (The WebSocket protocol) to Proposed Standard: request for max frame size
Hi, Recently, I posted [1] that websocket protocol should include an indication about max frame size that is willing to accept the connecting peer. Many pointed this is not an issue because you could use a stream oriented API (like TCP send/recv and others), but that only bypasses the problem (in some cases) not solves it. Websocket protocol is frame based and every frame based protocol designed until now has/need such feature or similar. Even TCP, for those that proposes to use a stream oriented API as a solution, includes a more complex mechanism than a simple max frame size (window based ack), so each party can control/inform the sender. This is critical. A good example for this would be the next. Let suppose you have a constrained memory device (either because it is small or because it accepts a large amount of connections). On that device you deploy a websocket app that only accepts small messages (< 512 bytes, let's say). In this context, you can hard code at your web app (javascript) that must not send messages larger than this size (so you control websocket payload size) and in the case you need to, you must split them properly. However, even with this mechanism, you have no warranty that the browser or an intermediary will join those messages into a single frame, causing your device to receive a bigger message/frame than expected. Assuming this I think we would require either to include a Max-Frame-Size indication (or a really poor solution: to explicitly state on the draft that frames must not be coalesced by intermediaries or browsers). Regards, [1] http://www.ietf.org/mail-archive/web/hybi/current/msg07641.html -- Francis Brosnan Blázquez ASPL 91 134 14 22 - 91 134 14 45 - 91 116 07 57 AVISO LEGAL Este mensaje se dirige exclusivamente a su destinatario. Los datos incluidos en el presente correo son confidenciales y sometidos a secreto profesional, se prohíbe divulgarlos, en virtud de las leyes vigentes. Si usted no lo es y lo ha recibido por error o tiene conocimiento del mismo por cualquier motivo, le rogamos que nos lo comunique por este medio y proceda a destruirlo o borrarlo. En virtud de lo dispuesto en la Ley Orgánica 15/1999, de 13 de diciembre, de Protección de Datos de Carácter Personal, le informamos de que sus datos de carácter personal, recogidos de fuentes accesibles al público o datos que usted nos ha facilitado previamente, proceden de bases de datos propiedad de Advanced Software Production Line, S.L. (ASPL). No obstante, usted puede ejercitar sus derechos de acceso, rectificación, cancelación y oposición dispuestos en la mencionada Ley Orgánica, notificándolo por escrito a: ASPL - Protección Datos, C/Antonio Suárez 10 A-102, 28802, Alcalá de Henares (Madrid). ___ Ietf mailing list Ietf@ietf.org https://www.ietf.org/mailman/listinfo/ietf
RE: [hybi] Last Call: (The WebSocket protocol) to Proposed Standard: request for max frame size
Hi Len, > I agree that this would be very useful. > > Would this be one frame size for both directions, or could it be specified > in each direction? It should be done in both directions, assuming each party may have different requirements.. > I'm a little wary of intermediaries being allowed to adjust this unless > they're only allowed to reduce the amount... My impression is that this max frame size value would help intermediaries, and peers, to know how to fragment or coalesce application level messages. Without such indication they are blind... ___ Ietf mailing list Ietf@ietf.org https://www.ietf.org/mailman/listinfo/ietf
RE: [hybi] Last Call: (The WebSocket protocol) to Proposed Standard: request for max frame size
> Francis, Hi Len, > I agree with your points and would welcome a max frame size negotiation > header. Ok, It shouldn't be a negotiation but an indication to accommodate communication to each party. > However, whilst an intermediary might legitimately change the fragmentation > of a frame it cannot merge complete messages. If, in your example, your > client limited itself to messages of a certain size Ok, it is important to note the peer is limiting max frame size not the message size which may be still infinite in practical terms (63 bits)and we have fragmentation support to deal with this. This way intermediaries can split/coalesce fragments according to receiving peer expectation. > then it doesn't matter > what intermediaries do to the frames as the maximum frame size would be > limited to the maximum message size that the client sends. The only time > this wouldn't work is if you were using a 'message of infinite fragments' > where you start with a fragmented frame and don't know when you'll send the > final fragment. Assuming frame size and message size aren't equal, it does matter. > The communication between peers about maximum message sizes supported could > just as easily be in the application level protocol that you're running over > the websocket connection. Ok, I see you point and in part you are right, but there is a risk with this argument: 1) Protocol problems should be solved on its layer, otherwise it will cause next layer (application level in this case) to need to solve them. It looks reasonable not forcing people to do so. Side note: our intention is to layer BEEP on top of websocket, so, in practical terms this is not a problem for us because it will provide us with all the protection and flow control required. However it looks to me this is not a solution for people using websocket in other contexts. 2) In general, it has lot of benefits to fragment bigger messages into smaller pieces to avoid sick interactions. For example, if you send a really large frame, you won't be able to send a control message in the middle until you finish sending the frame in place. Having a max frame size indication will help websocket stacks to fragment using the right value for each application. 3) No matter API style provided by a websocket stack (frame indication or stream oriented), having framing running in the background where the sending peer and intermediaries can coalesce frames without any indication of the upper level will cause problems that can't be solved in practical terms by a receiving side. In other words, I think having a framing without a max frame size mechanism (or other mech like ack confirmation, its the same) is like pretending having a coin with only one side. > Len > www.lenholgate.com > > > -Original Message- > > From: hybi-boun...@ietf.org [mailto:hybi-boun...@ietf.org] On > > Behalf Of Francis Brosnan Blazquez > > Sent: 12 July 2011 15:14 > > To: Hybi > > Cc: ietf@ietf.org; draft-ietf-hybi-thewebsocketproto...@tools.ietf.org > > Subject: Re: [hybi] Last Call: > > (The WebSocket > > protocol) to Proposed Standard: request for max frame size > > > > Hi, > > > > Recently, I posted [1] that websocket protocol should include an > > indication about max frame size that is willing to accept the > > connecting > > peer. > > > > Many pointed this is not an issue because you could use a stream > > oriented API (like TCP send/recv and others), but that only > > bypasses the > > problem (in some cases) not solves it. > > > > Websocket protocol is frame based and every frame based protocol > > designed until now has/need such feature or similar. Even > > TCP, for those > > that proposes to use a stream oriented API as a solution, includes a > > more complex mechanism than a simple max frame size (window > > based ack), > > so each party can control/inform the sender. This is critical. > > > > A good example for this would be the next. Let suppose you have a > > constrained memory device (either because it is small or because it > > accepts a large amount of connections). On that device you deploy a > > websocket app that only accepts small messages (< 512 bytes, let's > > say). > > > > In this context, you can hard code at your web app (javascript) that > > must not send messages larger than this size (so you control websocket > > payload size) and in the case you need to, you must split them > > properly. > > > > However, even with this mechanism, you have no warranty that > > the browser > > or an interm
Re: [hybi] Last Call: (The WebSocket protocol) to Proposed Standard
> As said before, making such DNS SRV specification an extension (so > present in other document) will mean no success at all, as WebSocket > client implementors (i.e. webbrowser vendors) will not be mandated to > implement it and service providers could not rely on the support of > DNS SRV in web browsers. So nobody will use them (because IE10 decided > not to implement it, for example). IMHO this is sad due the real > advantages DNS SRV provides for a protocol like WebSocket. > > Yes, in HTTP there is no special DNS stuff, all the load-balancing and > failover mechanism are done at server side with very complex and > expensive solutions (www.facebook.com resolves to a single IPv4 ). > The question is: should we also inherit every HTTP limitation in > WebSocket? +1 It is little the effort for web browser implementors and will provide lot of benefits to end users and developers which would be able to provide scalable/configurable services easily. I'm with Iñaki: its usage must be mandatory to really make it available in all browsers... ___ Ietf mailing list Ietf@ietf.org https://www.ietf.org/mailman/listinfo/ietf