I like your approach. Too many "performance enhancing" projects founder on not actually seeing what works and what doesn't. For some reason I thought there was already work on DIME and TCP in Axis2. But I've never actually looked at the code so I'm probably wrong.
Paul On 10/8/07, Dennis Sosnoski <[EMAIL PROTECTED]> wrote: > > Sure, I think that could be an excellent idea. > > On binary XML, the W3C EXI group (http://www.w3.org/XML/EXI/) is likely > to be the long-term winner. Sun's "Fast Infoset" has never seemed all > that fast in tests I've seen (including when used with Axis2), but if > you're looking for a short-term solution my own XBIS > (http://www.xbis.org) is a simpler alternative. When I added XBIS > encoding to my old JibxSoap code I more than doubled performance. Right > now I have it structured so JiBX can output directly to XBIS and read > directly from XBIS; if I extend this to include a StAX wrapper it should > be usable directly with Axis2. It'd be interesting to see how much > benefit that provided, vs. "Fast Infoset". > > On the TCP model, I'm currently implementing a simple DIME-based > approach for my own work. However, that does not allow for sharing the > socket connection (something that both Sun's approach and the .Net 3.0 > net.tcp technique support). I'm going to try some tests to see how much > benefit you actually get from sharing the socket connection. Offhand I'd > think the gains would be pretty small, but if they're significant that's > probably worth doing. DIME doesn't support that directly, but could > easily be extended to do so. > > - Dennis > > > Paul Fremantle wrote: > > I hate to suggest something new (NIH) but maybe we could start an open > > discussion and forum on what would make a good BinaryXML/TCP model? > > If we came up with something significantly better then it would be worth > > doing. > > > > Paul > > > > On 10/8/07, Dennis Sosnoski <[EMAIL PROTECTED]> wrote: > > > >> Hi Paul, > >> > >> I think it's an interesting possibility. I've been trying to find out > >> what .Net is currently using for their net.tcp transport, but that's > >> proving difficult. The Sun proposal is at least well-defined. > >> > >> I wish they'd asked for comments and started a discussion rather than > >> just making something up, though. Some parts seem lame, such as using > >> nibble encoding, and requiring a response message for every request > >> message (not necessarily appropriate with WS-Addressing involved). It's > >> also a bit heavy-weight, with a SOAP service request to open a channel. > >> > >> - Dennis > >> > >> -- > >> Dennis M. Sosnoski > >> SOA and Web Services in Java > >> Axis2 Training and Consulting > >> http://www.sosnoski.com - http://www.sosnoski.co.nz > >> Seattle, WA +1-425-939-0576 - Wellington, NZ +64-4-298-6117 > >> > >> > >> > >> Paul Fremantle wrote: > >> > >>> Guys > >>> > >>> We should develop a compatible transport to this transport: > >>> http://www.infoq.com/news/2007/10/soap-tcp-wcf > >>> > >>> Paul > >>> > >>> > >> --------------------------------------------------------------------- > >> To unsubscribe, e-mail: [EMAIL PROTECTED] > >> For additional commands, e-mail: [EMAIL PROTECTED] > >> > >> > >> > > > > > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > -- Paul Fremantle Co-Founder and VP of Technical Sales, WSO2 OASIS WS-RX TC Co-chair blog: http://pzf.fremantle.org [EMAIL PROTECTED] "Oxygenating the Web Service Platform", www.wso2.com
