Interconnect Software Consorium Socket API Extensions Working Group. -------------------------------------------------------------------- Meeting: Thursday 7th March 2002 21:00GMT (04:00p EST, 03:00p CST, 01:00p PST, 22:00CET, 06:00JST+) Toll Free (US & Canada): (877) 214 5010 International: +1 (504) 588 9772 Code: 769 497 Host (needed to kick things off): 237 426 -------------------------------- Present: David Edmondson, Sun Masanori Itoh, Fujitsu Jeremy Harris, Sun John Kasperski, IBM Annie Foong, Intel Viswanath Subramanian, HP Louis Laborde, HP Jack McCann, Compaq Jeff Messing, IBM Juliana Hsu, IBM Satya Sharma, IBM Carl Hensler, Sun -------------------------------- Minutes: 1. Agenda bashing. None. 2. Note new teleconference number for next week. (dme) See at the tail of this document ! 3. Review action items. See below. 4. Timescales. (dme) See (very) recent email from the SC on proposed timescales. No disagreement with the timetable was expressed. Carl described how it's okay to change the schedule somewhat, but that has to be done after some thought. The first milestone (April 2002) would be an appropriate time to indicate that the proposal is or is not viable. 5. Requirements specification. (dme) How are we going to produce it ? The group discussed approach. Satya and David will produce an outline document and start to fill in the gaps. Others will be asked to contribute to fill in relevant areas. Everyone is encouraged to send proposals for items to be included in the requirements to the mailing list. ACTION: dme - Talk to Martin Kirk about document formatting. Carl: have a look at the SDP requirements for a sample. Jeff: remember to note where a requirement came from in order that we can go back and discuss it again with the relevant party. 6. Areas of interest discussions. Masanori: support for user-level protocol implementations, where (for example) file descriptor passing is insufficient as a mechanism for communicating state between Unix processes. There would appear to be a trade-off of performance vs. semantics for user-level protocol implementation. Satya: kqueue. Currently provides a notification framework, but not the async operations themselves. Jeff: Concerns about the correctness requirements (i.e. pending notifications should be delivered at the time interest is registered) - could be difficult to implement correctly and efficiently. Jeff: Issues about memory requirements of the API as currently specified, as each notification would appear to require allocation. Jeff: The ability to aggregation notification events, but retain the knowledge that they were originally distinct events is interesting. Satya: IOCP's correlator implies a 1 to 1 relationship between issued operations and notifications. Juliana: Winsock allows you to wait for one of N events. General conclusion is that we should probably not simply adopt kqueue "as-is", but that it has many interesting properties which are worth considering. ACTION: Jeff to extract interesting requirements from the kqueue paper and forward a list we hope to leverage to the list. ACTION: dme to do the same for /dev/poll implementation and analysis. ACTION: dme to send linux aio notes pointer. Satya: Buffer pinning/memory management is an area that doesn't appear to have a lot of published prior art. Need to cooperate with the native group, as they will also be looking at this. ACTION: Satya talk to native group about buffer pinning. 7. AOB. We need more items added to the agenda in advance in order that we can drive the meeting more effectively. Please send suggestions to the list or co-chairs. -------------------------------- Outstanding Actions: 2002-02-28/01 (dme) Check with the steering committee for any hints or guidelines on the form of the requirements specification. 2002-03-07: mail sent. See the timescales returned by Martin Kirk and the suggestions to an approach from Mike Krause. Carl: don't get bogged down in the requirements document. Carl: order the requirements in priority, ensure more detail on the higher priority items. DONE: 2002-03-07 2002-02-28/02 (dme) Talk to the steering committee and IETF application area directors about possible cooperation. 2002-03-07: mail sent to icsc-sc. Carl: informal cooperation is fine. Remain aware of the work already underway, but don't be distracted. The ICSC is the place where definitive work will be done. DONE: 2002-03-7 2002-02-28/03 (Satya) Get the IBM (DB2 ?) people added to the mailing list. DONE: 2002-03-07 2002-03-07/01 (dme) Talk to Martin Kirk about document formatting. 2002-03-07/02 (Jeff) Produce a summary of requirements extracted from the kqueue paper and send it to the list. 2002-03-07/03 (dme) Product a summary of requirements extracted from experience with /dev/poll. 2002-03-07/04 (dme) Send a pointer to the Linux AIO notes paper to the list. DONE: 2002-03-07 2002-03-07/05 (Satya) Open a dialogue with the Interconnect Transport API working group about buffer management. -------------------------------- Next Meeting: Thursday 14th March 2002 21:00GMT (04:00p EST, 03:00p CST, 01:00p PST, 22:00CET, 06:00JST+) Toll Free (US & Canada): 1-888-742-8686 International: +1-303-928-2600 Code: none required -------------------------------- Contact Details: email: icsc-socketwg@opengroup.org web: http://www.opengroup.org/icsc/sockets/