Nsis_Imp January 2005

nsis_imp@informatik.uni-goettingen.de

12 participants
19 discussions

Start a nNew thread

Building blocks
by Xiaoming Fu 13 Jan '05

13 Jan '05

Hi all, Attached please find some basic modules (draft) of a simplest GIMPS prototype impl... Cheers, Xiaoming

1 0

Cleo meeting minutes 13.01.05
by Xiaoming Fu 13 Jan '05

13 Jan '05

[XiaomingFu] all our chat info are available (sounds not "private"): http://www.jabber.org/muc-logs/cleo-project.org@private.jabber.org/2005-01-… http://www.jabber.org/muc-logs/cleo-project.org@private.jabber.org/2005-01-… [13.01.05 13:51] elwynd Away as a result of idle. [13.01.05 13:51] elwynd Available [bschloer] AW: How's your ssl-implementation moving forward? [awester] BS: not much, cause i'm currently working on some other thins in parallel. [bschloer] AW: I see [XiaomingFu] AW provided recent description about his progress: http://cleo-project.org/?DevDiaryAw [elwynd] i'm back [13.01.05 13:54] julabeille Présent [bschloer] AW: You mentioned to blocking and non-blocking behavior of accept. maybe can talk about this after the chat? [awester] BS: yes, of course. [bschloer] hi julien [julabeille] Hi bernd and everybody [awester] hi [13.01.05 13:56] youabidi Online [julabeille] hi Youssef [youabidi] Hi julien, hi everybody...! [bschloer] hi youssef [XiaomingFu] hi all [XiaomingFu] it's fast now [bschloer] hi xiaoming [XiaomingFu] do we wait for others (mainly, SW & HT) [awester] hannes will not join us today. [XiaomingFu] CA is unable to join today [elwynd] ht told me he will not be attending as he has a meeting with his boss [bschloer] we could wait until 14:15 and start then, what do you think? [XiaomingFu] AW: did you talk with Hannes about next open tasks? [awester] XF: he told me there's something to do for me related to gimps states. [XiaomingFu] would you be happy with it [awester] XF: yes, this task doesn't sound very problematically to me. [13.01.05 14:05] 11sliN11 Online [julabeille] is there a page on cleo-project related to todays' meeting? [13.01.05 14:06] 11sliN11 Offline [13.01.05 14:06] nroettger Online [nroettger] hi all [XiaomingFu] hi Julien - there is a "new open tasks" page [XiaomingFu] All: Once today's meeting, we ("steering committee" ) will be only observing your progress (e.g., cleo page & bi-weekly team meeting) and providing help if you raise for us. the real activities and information exchange will take place within your team folks [julabeille] ok [bschloer] JA: no, i don't think so. we're discussing mainly about the new 'open tasks' [bschloer] hi nils [bschloer] for the new entered people: we want to wait until 14:15 to let other people enter and start then [XiaomingFu] BS - Is sebastian also joining today? [bschloer] XF: I haven't he hasn't told me [julabeille] BS: some administrative questions are not resolved : i think of syntax used in code, i just saw updates on the related page in "discussions", i am ok for any syntax but it would be nice to decide (and it does not take much time) [nroettger] XF: yesterday he said he will join us [XiaomingFu] Elwynd told me he will investigate the XORP state machine support and reported back to the team [XiaomingFu] shall we wait for him a little bit [bschloer] JA: Can you short say, what's unsolved? [bschloer] XFure [nroettger] BS syntax used in code [julabeille] decide between GNU and linux style [julabeille] i meant decision [bschloer] I just wanted to read the relating pages: I get 'page not found'. Are there special passwords? [nroettger] no [julabeille] did you try "discussions -> syntax convention for the ... ? [bschloer] JA:yes. and there are two subitems: Linux & GNU style [nroettger] linux style: http://pantransit.reptiles.org/prog/CodingStyle.html [bschloer] JA: strange. now i can access them [julabeille] BS: so do you have any preference ? [bschloer] JA: no, i don't have any prefs. [bschloer] to the group: which style do you prefer? [nroettger] I don't care [julabeille] JA: no pref [awester] i prefer the linux style [youabidi] I prefer linux too... [youabidi] (But I don't really care either...) [awester] [bschloer] ok, is everybody happy with the linux style? [elwynd] i used to prefer the gnu style but i will live with either [bschloer] I'd like to welcome everybody for the chat [bschloer] a lot of things changed quickly and there's not much time to proceed [bschloer] mainly there are two or three items i want to talk about: [bschloer] 1 the new jobs have to be distributed to the group [bschloer] 2 talk about the time-frame [XiaomingFu] If you don't mind for my oppion, I would suggest to quickly decide Linux style (as it seems to most popular among the team) [bschloer] 3 some short issues about communication [julabeille] XF: ok for linux style [bschloer] XF: linux is ok [bschloer] let's start with the new work [bschloer] there are: session storage/representation, state machine, socket-management and event-loop [bschloer] i'd like to ask the group, if there're any preferences. does anybody like to take one of these jobs? [elwynd] BS: as we discussed just now i wil look at xorp state machine stuff and kick off state machine stuff if you like [awester] XF: i think the session storage/representation stuff belongs to me, am i right? [bschloer] EW & AW sounds perfect perfect to me [XiaomingFu] AW: it's up to your preference/decision, anyway - just a personal suggestion [bschloer] There are 2 items left: socket-management and Event Loop [awester] BS: so i'll take session storage/representation. [bschloer] I'd say I continue with the event loop stuff [bschloer] the socket management represents: udp/tcp - -router alert [XiaomingFu] i.e., how to interact with sockets (tcp, udp & raw) [bschloer] XF: do you have any idea who could do this? [julabeille] Youssef and I have to finish the parser, which will last ? time, then we will have muck more time (studies finish in a week ffor me) [bschloer] JA: so you could take that part? [elwynd] i can provide support for socket stuff if people feel they need bootstrapping [julabeille] BS: my timing is : (1 week of studies - but i will have quite a bit of time, then 3 weeks totally available) , if you think i can do profitable things in that time, i will take the socket part [XiaomingFu] fyi: XORP seems provide raw socket API as well: http://www.xorp.org/releases/current/docs/fea/fea.pdf [bschloer] sounds great. [bschloer] NR: we discussed to continue writing test-cases [bschloer] NR: they are not needed right now but in the next part [youabidi] BS: Maybe I can join julien in the socket part. [youabidi] I know that TCP is not *really [youabidi] * complicated but UDP is something else. [julabeille] thank you Youssef !! [bschloer] YA: Yes sure, i thought this as well as you were working together [youabidi] Perfect. [bschloer] the next issue is the time frame: [elwynd] YA: UDP is in many ways much simpler! [bschloer] there exactly 7 weeks left until the ietf meeting which is on the 06.03.05 [youabidi] EW: That sounds reassuring...! [awester] BS: I'm very busy the next two weeks, because I've to write some exams. [nroettger] BS: me too [bschloer] how much time can everybody spend on this project the next 7 weeks? [awester] BS: 7,5 hours/weekday. [elwynd] BS: 1-2 days per week [julabeille] BS: 10 h next week + 3 entire weeks [youabidi] BS: At least 4 weeks, about 15h/week. [bschloer] I can work full-time beginning from next Tuesday [nroettger] BS: 10h/ week, perhaps more after the first exams [bschloer] ok, i'd like to focus on the 20.02.05 [bschloer] this is 5 weeks to go and then there are 2 weeks buffer [awester] this sounds perfectly to me! [awester] perfect [bschloer] please use the cvs and the developer's diary [bschloer] to the group: what do you think, how often we should chat: once/week - 2 x /week - every second week? [bschloer] i'd like to setup a platform/wiki on the cleo-project.org to discuss certain things as wekk [bschloer] as well [nroettger] I think we have to use the homepage more often, the we can meet every two weeks in the chat [julabeille] i am not too used with that, but i spoke with CA about "fonctionnal blocks", which detail the tasks and persons working on them, i would be happy if you could do something like that to have a global view of the project [julabeille] (BS) [XiaomingFu] I can draw a figure about the overall blocks shortly [julabeille] Bs : once a week seems good to me, completed by 1 to 1 chats when needed [bschloer] ja & xf: a picture of a functional block-diagram is good for an over-all understanding [youabidi] BS: I agree with the fact that the homepage is good place to put the evolutions of the projects. Meetings should IMHO only be used to make important decisions. [youabidi] BS: I agree with JA. Once a week sound OK. [bschloer] ok, we'll see what progress is made and we can can do the chat chat when needed [bschloer] let's focus once a week [awester] BS: I align myself with JA and YA. [julabeille] YA, BS : i totally agree to use the home page, so that chats do not go in too many directions [bschloer] this was also the third issue i wanted to talk about: communications [bschloer] that's all i wanted to talk to you. [bschloer] does anybody of you have any questions or issues? [nroettger] bs: what do you mean with: test will be needed in the next part? [julabeille] BS : for technical and not general topics, i used to contact Cedric or Sebastian once a week aproximately and it was also very profitable to me [julabeille] (communication issue) [bschloer] NR: to finish this project until begin of march we want to keep the project as simple as possible to make it run [bschloer] nr: so there will be not much testing [bschloer] nr: sebastian told me, the test functions will be needed in the further progress of the project [bschloer] JA: you can contact me whenever you want or when there are problems, no problem [julabeille] BS : ok, great [bschloer] of course the whole group can contact me [bschloer] ok. let's move over to the project-home page. I'll do some changes to setup a communication platform [bschloer] and please feel free to contact me with suggestions or comments [bschloer] i'm quite new to the project and i'm very happy about help [bschloer] a good point for the diary would be to start like: what have i done since last week/time, what problems appeared and what i'll to until next week/time [bschloer] ok, that's it what i wanted to say. [youabidi] This is okay for me (an JA). [youabidi] (That may even be the only way for us to make others aware of what is done...) [bschloer] if there are no additional comments i don't want to waste your time [XiaomingFu] it's great to see fast progress. i've no comments [bschloer] xf: thanks [elwynd] bs: thanks.. no comments [bschloer] thank you for taking part in this meeting. good luck ! [julabeille] thank you eveeribody [youabidi] Thank you everybody...!

1 0

Re: [Nsis_imp] Nonination of Bernd Schloer to become CLEO team leader]
by Bernd Schloer 11 Jan '05

11 Jan '05

Hi list, thank you for your confidence in me. I will do my best to make this project run until begin of March. I suggest that we meet this Thursday at 14:00 in the jabber chatroom where we met last week. Bye, Bernd > > Hi all, > > and welcome Bernd, > any time after 12:00 is ok for me. > > Best regards, > Julien > > Xiaoming Fu a écrit : > >> Hi all, >> >> Due to Sebastian's resign, Cedric, Hannes and myself have discussed >> about nonimating a new team leader, where we believe Bernd Schloer >> would be the best people for this role. Welcome Bernd! >> >> Next I'd propose to have a project meeting on Thurday to clarify the >> necessary stuffs, such as project organizations and subtasks as much >> as possible, in order to avoid confusing and inefficiency in later >> development. Any time before 16:00 would be ok for me personally. >> Please make suggestions. >> >> We thank again for Sebastian's critical contribution and believe he >> will continue to help moving our team towards a success. Moreover, we >> look forward to all contributions from all team members! >> >> Best regards, >> Xiaoming >> _______________________________________________ >> Nsis_imp mailing list >> Nsis_imp(a)informatik.uni-goettingen.de >> http://user.informatik.uni-goettingen.de/mailman/listinfo/nsis_imp >> >> > > > > _______________________________________________ > Nsis_imp mailing list > Nsis_imp(a)informatik.uni-goettingen.de > http://user.informatik.uni-goettingen.de/mailman/listinfo/nsis_imp >

2 1

Some thoughts about the past and future of cleo-project.org
by Sebastian Willert 11 Jan '05

11 Jan '05

Hi all, I think it is high time after the confusion of the last week to get this project on track once again. I regret that I had to come to the decision to step down as project leader but I still think it is in the best interest of the group. I want to sincerely thank all of you for the support and encouragement you have given me in the last days and the good collaboration since the start of cleo-project.org, but now it is time to work out how this project will continue. I have started with the false premise that Cedric, Hannes and Xiaoming have no time to get involved into this project but stand ready to support us when their help is needed. This was a misconception on my part and lead to the current state of affairs. I think the main problem with our former approach were the three layer hierarchy and the lack of communication between me and the steering committee. In my opinion the only short term solution was to remove one level, and as I am the only one that could be removed I decided to step down. I wished I had taken these problems more seriously from the beginning, but I have to accept that I failed in this respect and have to accept the consequences. If this had been noticed earlier, it would have been possible to seek a solution for the challenges posed by such a set-up, but now time is running too short to work out a good solution. I have the feeling that Cedric et al. are more than capable of handling this project well and I offer you all the support I can give, but I cannot accept the responsibility of leading this project at this stage. Regretfully, Sebastian P.S. Please don't get de-motivated by these events. I still think that it is possible to reach our goals and you still have my full support in this pursuit. I just had to readjust my position for the sake of the team. Maybe I'll even start to contribute some code now that I have some time at hands :)

3 2

Notice of resignation
by Sebastian Willert 07 Jan '05

07 Jan '05

Hi all, I regret that I have to let yo know, that I decided to hand in my resignation. I have the feeling that there are insurmountable differences in collaboration philosophy between me and the steering committee. Yesterday was a sad day for me: the meeting that would have been very important for my next steps turned into absolute chaos and I owe you all an apology that I couldn't keep myself out of the arguments that destroyed the meeting. I am sure that my leave is to beneficial to the project: I have no problem offering my help from the sideline and maybe I can help you more when I am not burdened with all the management overhead and power struggles. Regards and good luck to all of you, Sebastian

2 1

meeting minutes 6jan2005
by Tschofenig Hannes 07 Jan '05

07 Jan '05

hi all, here are the meeting minutes from our yesterday's discussion: ---------------------------------------------------------------------------- --------------------------------------------------------------- [2:05 PM] HannesTschofenig hat den Raum betreten. [2:05 PM] cleo-project.org [2:05 PM] awester hat den Raum betreten. [2:05 PM] bschloer hat den Raum betreten. [2:05 PM] elwynd hat den Raum betreten. [8:12 PM]<xiaomingfu> thanks [8:12 PM]<elwynd> that is 14.00 CET? = 13.00 GMT [8:13 PM]<julabeille> yes [8:13 PM]<elwynd> ok.. ttfn (ta ta for now -= bye) [8:14 PM]<xiaomingfu> good evening [8:14 PM]<julabeille> good evening [8:14 PM]<youabidi> Good bye everybody...! [8:14 PM]<bschloer> bye, cu then [8:14 PM]<HannesTschofenig> bye. [8:15 PM]<s.willert> good night everybody [8:16 PM]<awester> bye [8:17 PM]<bschloer> SW: i sent you a private message [8:18 PM]<N.Roettger> bye all [8:18 PM]<s.willert> xf: can you aithorize me so I can add you to my buddy list? [8:19 PM]<xiaomingfu> yes, how [8:19 PM]<s.willert> BS: sorry, I haven't got it.... just IM me [1:07 PM]<bschloer> hello ! [1:07 PM]<HannesTschofenig> hi. [1:07 PM]<HannesTschofenig> didn't we want to start at 14:00? [1:07 PM]<bschloer> let's wait for the others. [1:08 PM]<bschloer> yes, 14:00 [1:08 PM]<HannesTschofenig> do you experience the same latency? when i send a message then it takes a few seconds until my message appears. [1:09 PM]<bschloer> sometimes yes, but most of the time my message appears without latency [1:09 PM]<HannesTschofenig> interesting. maybe i should setup my own jabber server.... [1:20 PM]<bschloer> ok, cu @ 14:00 [1:52 PM]<bschloer> hello again [2:01 PM]<elwynd> hi [2:02 PM]<awester> hi! [2:02 PM]<elwynd> is sebastian joining? [2:04 PM]<11sliN11> hello [2:05 PM] N.Roettger hat den Raum betreten. [2:07 PM] XiaomingFu hat den Raum betreten. [2:07 PM]<XiaomingFu> hi [2:08 PM]<XiaomingFu> what's the missing point will be the topic today? [2:09 PM]<elwynd> when sw arrives address the remainder of his points from the web... doc methods, collaboration file owbnership strategy, etc [2:10 PM]<XiaomingFu> I think he will be here [2:15 PM] julabeille hat den Raum betreten. [2:15 PM]<bschloer> elwynd: hi! what's your real name? i'm updating my buddy list [2:15 PM]<julabeille> bi [2:16 PM]<elwynd> in the meantime did anybody have any more thoughts on the issues we discussed yesterday? NTLP/NSLP interface, session repreesentation, XORP updates? [2:16 PM]<XiaomingFu> let's start the discussion without him [2:16 PM]<elwynd> bs: Elwyn Davies [2:16 PM]<XiaomingFu> him --> SW [2:16 PM]<XiaomingFu> when he is here, we can confirm to him [2:18 PM]<julabeille> maybe it would be good to evoque the different points prepaired by Sebastian on the web site [2:18 PM]<elwynd> ok... any input? [2:18 PM]<julabeille> sorry, input? [2:19 PM]<elwynd> anu body got anything to say about either any of yesterday's topics or documentation methods? [2:19 PM]<XiaomingFu> ==DiscSourceCodeDocumentation - doxygen (more or less the standard way of documenting c and c++ sources) - XORPs version of xdoc (The XORP project uses a modified version of the KDE documentation tool kdoc tool to produce source code documentation) == DiscDesignDocumentation - latex (prooven, flexible, confusing, beautiful, and far from easy to learn) - docbook (xml-based documentation engine, not as flexible as latex, but this might be a Good Thing) - HTML (yack!) [2:20 PM]<XiaomingFu> still, would we seek for options listed in http://cleo-project.org/?DevMeetings [2:20 PM]<HannesTschofenig> i haven't seen a document (description) which tells how the software architecture looks like. [2:20 PM]<HannesTschofenig> i haven' t read the xorp docs and therefore i do not know what the options are. [2:21 PM]<HannesTschofenig> but i think such a document is need soon. yesterday we had a discussion about the one process vs. multiple process, about apis, etc. [2:21 PM]<HannesTschofenig> these issues obviously depend on each other. [2:21 PM]<HannesTschofenig> does xorp also provide some mechanisms to handle multiple session at the same time. [2:21 PM]<elwynd> xf: do you want to summarize our discussions of this morning for ht [2:22 PM]<HannesTschofenig> i know some other apis which provide such an abstraction. [2:22 PM]<HannesTschofenig> discussions this morning? i missed something? [2:22 PM]<XiaomingFu> software arch: option 1: single process, multi-thread, callbacks: http://user.informatik.uni-goettingen.de/~casp/ICNP04_casp.pdf option 2: scheduling using xorp, incl. state machines and event loops [2:23 PM]<HannesTschofenig> i see. [2:24 PM]<HannesTschofenig> you have used threads to deal with multiple concurrent sessions - as a replacement for select. [2:24 PM]<XiaomingFu> HT: we (ED & CA and me) discussed about the design appraoch this morning: up-to-down or down-to-up, we reached a consensus to follow the latter, i.e., the current approach that SW is trying to push to [2:25 PM]<HannesTschofenig> very good. so what is the proposal? [2:25 PM]<elwynd> at least in the first fast prototype... to be ready for next IETF [2:25 PM]<HannesTschofenig> that's a good point. we don't need to get best result at first. we can improve later. [2:26 PM]<XiaomingFu> HT: before we can calibly make the code spits we will have several trials and func testing with xorp and see how we can incorporate protocol in xorp [2:27 PM]<elwynd> so the idea is to do a bit more investigation... discovery.. to see how xorp scheduling/callbacks will work as i understand [2:28 PM]<XiaomingFu> HT: the overall consensus this morning was: we four (ED, CA, you & me) focus on providing protocol assistance when necessary, while SW and the dev. team focus on system level exploration, design and interaction. try to provide a prototype-of-prototype in Feb, and towards a pre- release in IETF meeting [2:29 PM]<HannesTschofenig> sounds good to me. [2:29 PM]<HannesTschofenig> i personally do not like to have a too strong coupling with xorp since the software should run on a regular linux box as well. [2:30 PM]<HannesTschofenig> these libraries are cool but they always have some drawbacks. i know this from the ace library. [2:30 PM]<bschloer> HT: what do you mean by drawbacks? [2:31 PM]<HannesTschofenig> typically these libraries do not work well with security protocols (e.g., tls). furthermore, it takes a long time until you understand these libraries. typically there are some bugs which are really, really hard to find. [2:31 PM]<awester> the xorp libraries should support most linux versions with kernel 2.4.x and later, refering to their internet site. [2:31 PM]<HannesTschofenig> i see. that's good news. [2:33 PM]<HannesTschofenig> keeping things simple for the first version with the capability to add more sophisticated functionality later on. [2:33 PM]<elwynd> the good thing about xorp is that it is being actively developed by people with grants!... so maybe bug fixes will happen [2:33 PM]<HannesTschofenig> even the multi-threaded stuff is too much for the first version. the performance improvement of a multi-threaded version compared to one which uses select is possibly minor. [2:34 PM]<XiaomingFu> certainly, we don't care for the first version - as long as xorp is found to be easy & how to handle IPCs and scheduling [2:35 PM]<elwynd> as i said yesterday, it is debateable whether a fully asynchronous multi-threded version is actually desirable... presents difficukties in load control [2:37 PM]<elwynd> Is that done then? move on to documentation? [2:37 PM]<XiaomingFu> berd: did you manage to succeed in testing xorp's func? [2:39 PM] youabidi hat den Raum betreten. [2:39 PM] youabidi hat den Raum verlassen. [2:40 PM]<bschloer> XF: yes, but i found some weaknesses [2:40 PM]<XiaomingFu> e.g.? [2:40 PM]<bschloer> XF: e.g. i've problems to flush buffers [2:40 PM]<bschloer> XF: or to add new callbacks [2:41 PM]<HannesTschofenig> i once read the literature on select, poll and multi-threaded version. i couldn't find a clear argument why you should use multi-threaded servers (except if the message progressing is very high for a single incoming message). [2:41 PM]<bschloer> XF: existing callbacks are no problem [2:41 PM]<XiaomingFu> HT: i am not addicted to multi-threaded [2:41 PM]<XiaomingFu> BS: existing callbacks defined in the sample? [2:42 PM]<bschloer> XF: and i've a problem to access variables within the xorp-functions [2:42 PM]<XiaomingFu> local variable? [2:42 PM]<bschloer> XF: no, callbacks which are setup during program startup [2:42 PM]<bschloer> XF: yes, local variables [2:43 PM]<XiaomingFu> BS: but with regard to socket and event handling are ok? [2:43 PM]<bschloer> XF: one problem was to add a new callback after the accept-function, i solved it with dup (duplicating file descriptors) [2:45 PM] youabidi hat den Raum betreten. [2:45 PM]<bschloer> socket has one weakness: xorp tries to read from the fd after the event. even when the connection is not established (->before i can call accept) [2:45 PM]<XiaomingFu> what about presentation of state machine in xorp? [2:45 PM] youabidi hat den Raum verlassen. [2:45 PM]<julabeille> hi Youssef...bye [2:46 PM]<XiaomingFu> has it been tested? or assigned to be tested [2:47 PM]<bschloer> XF: i think SW has tested the state machine? [2:47 PM]<XiaomingFu> BS: I don't know the status - maybe you knew sth [2:48 PM]<HannesTschofenig> can someone point me to an xorp document which shows the functionality relevant for this discussion? (scheduling, event loops, etc.) [2:48 PM]<bschloer> HT: i have to look for one, i have my knowledge from the source-code [2:49 PM]<XiaomingFu> i think: http://www.xorp.org/design_docs.html [2:51 PM]<bschloer> XF: is the state machine the same as the event loop? [2:52 PM]<XiaomingFu> it is about automatic understanding of a (mostly text-based) array ==> easy translation of a defined state transitions. typically easier than manually in event loop. [2:52 PM]<XiaomingFu> but i don't know how xorp deals with it [2:54 PM]<bschloer> XF: so far i've experiance with the event-loop. the event-loop is based on the C-systemcall select. the effects are more or less the same. [2:56 PM]<bschloer> XF: with 'select' you can manage a couple of file discriptors (files, keyboard, sockets, ...). xorp does the same. [2:57 PM]<XiaomingFu> yes i know what do you mean [2:58 PM]<XiaomingFu> libxorp can be used for basis of the scheduler, event-loop, state-machine [2:59 PM]<bschloer> XF: you could show me the states, you want to implement. maybe then i can tell you more [3:01 PM]<XiaomingFu> e.g., any of GIMPS state machine http://user.informatik.uni-goettingen.de/~fu/nsis/statemachine/ntlp/draft-fu -nsis-ntlp-statemachine-00.html [3:02 PM]<HannesTschofenig> tseno has put the most recent state machine draft at: http://www.lrz-muenchen.de/~tseno_tsenov/gimps/ [3:02 PM]<HannesTschofenig> (visio/pdf files). [3:03 PM]<XiaomingFu> HT: probably the main parts we're going to be based on are: libxorp, xipc and libcomm [3:03 PM]<XiaomingFu> yes, the state machine is going to be resubmitted soon [3:04 PM]<XiaomingFu> just choose one example (table). I know in some protocol implementations there are such tables as text file and the code just reads the table and interprets them [3:05 PM]<HannesTschofenig> please send comments to the state machine soon. tseno is looking for feedback. [3:05 PM]<elwynd> i'll take a look as well [3:05 PM]<bschloer> HT: ok [3:06 PM]<HannesTschofenig> regarding the implementation of a state machine: often it is a table with pointers to functions. for eap and eap methods i have seen state machines as if then else statements. [3:06 PM]<HannesTschofenig> (or switch case) [3:06 PM]<XiaomingFu> yes, i agree (the table way) [3:08 PM]<elwynd> there are some clever tricks you can play with state machines described in Python using an object to maintain state [3:08 PM]<HannesTschofenig> interesting. can you point us to this document? [3:09 PM]<elwynd> yes... i'll find the ref [3:09 PM]<bschloer> XF & HT: i had a quick look at gimps_node_intermediate_clearing_transitions.pdf. so far i have not seen any implementation within xorp. but i have a close look on it now. [3:11 PM]<HannesTschofenig> what do you mean? you think that there the gimps state machines do not require any interaction with xorp libs? [3:12 PM]<XiaomingFu> http://www.xorp.org/releases/current/docs/kdoc/html/bgp/peer_hh.html [3:12 PM]<elwynd> ht: python state macines using generators: start form here - http://www.python.org/peps/pep-0255.html - there is a series of articles (or just search for 'generators state machines' at www.python.org [3:15 PM]<bschloer> HT: from the eventloop there are different event-handlers called. you could implement an own state-machine without xorp-libs. [3:17 PM]<XiaomingFu> can python approach be reused, or we impl. our own presenation of state machines? [3:17 PM]<HannesTschofenig> i need to take a look at the bgp code. [3:22 PM]<elwynd> xf: python approach relies on capabilities in pyton 2.4 but the idea of instantiating a class to hold the state of each connection is useful.. the python approach actually carries 'code context' through the use of iterators/generators.. reduces overhead but no direct equivalent in C/C++ [3:25 PM]<elwynd> are we done for now? or does anybody wish to express a view on documentation? [3:25 PM]<XiaomingFu> BS: i think you're right. they are equivalent to each other. event loop + callbacks [3:27 PM]<julabeille> Sorry guys, i have to go... about design questions, i added a short page in ?discussions about a convention for variable names in the code...comments are welcome [3:28 PM]<HannesTschofenig> the bgp stuff has a lot of code but the state machine looks nice to me. fairly easy to read. [3:28 PM]<julabeille> good afternoon [3:28 PM] julabeille hat den Raum verlassen. [3:29 PM]<XiaomingFu> julien - thanks for your efforts [3:30 PM]<XiaomingFu> too slow [3:32 PM]<bschloer> ok guys, i'd like to say good bye also ! [3:33 PM]<N.Roettger> bye bernd [3:33 PM]<awester> bye, bernd. [3:34 PM] bschloer hat den Raum verlassen. [3:34 PM]<XiaomingFu> Design doc: docbook? [3:34 PM]<XiaomingFu> too slow [3:36 PM]<XiaomingFu> slow speed because of this server? [3:37 PM]<elwynd> xf: I would think that is more future looking: on the code side it is less clear... my last project was happy with doxygen but if this is going to go back into XORP, their scheme would be more sensible...probably nearly equivalent? [3:37 PM]<elwynd> y...e....s [3:42 PM]<HannesTschofenig> i have a proposal: we have a few design decisions to make. it would be good if everyone would think about their favorite choices. what do you think? [3:42 PM]<XiaomingFu> i have no objection to any tool as long as the developers find any manipulatable [3:43 PM]<elwynd> ht: i second xf's view [3:47 PM]<XiaomingFu> i think the most important is to continue to discover the capability of xorp, esp. IPC, state machine, event loop [3:48 PM]<elwynd> ok. have to drop off now.. keep us posted on email please as investigations progress. [3:48 PM]<elwynd> bye [3:51 PM] elwynd hat den Raum verlassen. [3:52 PM]<HannesTschofenig> folks, i guess we should finish at this point. ---------------------------------------------------------------------------- --------------------------------------------------------------- ciao hannes

1 0

Todays meeting
by Sebastian Willert 05 Jan '05

05 Jan '05

Hi all, I've written up a few thoughts about things I'd like to discuss today in the wiki[1], please have a look and add your own items so anyone can prepare up front. I hope to establish such an agenda and publish it a few days before the next meetings, so stay tuned. Cheers, Sebastian --- [1] http://cleo-project.org/?DevMeetings

1 0

XORP-Library, AsyncFileReader-Object
by andreas＠overturn.de 05 Jan '05

05 Jan '05

Hi all! First, a happy new year to everybody! I got the chat application from Bernd and noticed a problem while reading through the source: The connection ist set up and teared down every time a message is sent from the client to the server. The accept() call takes place in the AsyncFileReader callback function associated with the server socket. Bernd told me that otherwise the application won't work if the connection isn't set up and closed before the callback-function returns. I think this behaviour is very inefficient, especially using SSL/TLS the cost of the handshake will be immense. Trying by myself to associate the AsyncFileReader object with the socket returned by a call to accept() and starting the event-handler after the connection has been established results in the following error message: > TestProject: /usr/local/src/xorp-1.0/libxorp/asyncio.hh:113: AsyncFileOperator::AsyncFileOperator(EventLoop&, int): Assertion `fl & 04000' failed Thinking on the next steps implementing the GIMPS layer, this will become more problematic: An important design decision of GIMPS is reusing existing connections and minimize setting up new ones. But how can this be done using this stuff? Any suggestions? Greets, Andreas -- eMail: andreas(a)overturn.de ICQ: 176261447 Jabber: awester(a)jabber.org

2 1

Status update for the cleo project
by Sebastian Willert 04 Jan '05

04 Jan '05

Hi all and welcome to our mailing list. The last weeks have been quite chaotic and I haven't managed to give you the feedback you deserve so I'll try to make up for this with this mail. First of all I am very happy to have you on board and I feel that we are starting to build up the momentum we will need to meet the March deadline. Most of our infrastructure is finally in place and all of you (at least the developers) should have received ssh, cvs and wiki accounts. The ssh accounts will give you access to a full fledged testing environment later on. This testbed consists of eight machines running IPv4, IPv6, SCTP, TCP, UDP, possibly NAT and Firewall environments and everything we need to test out our implementation. The cvs is nearly empty right now, but I will start committing our build framework and the adopted xorp libraries right after new years eve. You might want to check out the private project space[1]. You are encouraged to commit everything you've done for cleo into this repository, so everyone can have a look into your sources and it will be easier to discuss your questions and the problems you've encountered. I'd also like to invite you to take an active part in providing content on the website, not only for potential users but also, and in this phase more importantly, to provide feedback to your fellow developers. It is vital for this project that we communicate and keep the sense of buzzing activity alive, that I have witnessed during the last weeks. Keep us up to date on what you are doing right now, what problems you encountered or maybe just what music you are listening to while coding ;) I hope the developer diaries can make up for the physical distance between us and create a community out of loosely coupled developers. I'd also like to establish a two-weekly meeting on jabber (Cedric already started with this on Monday) to discuss the happenings of the time since the last meeting and our future plans. The first meeting is scheduled for Tuesday, January 4th at 1400 CET (Paris, Berlin). I hope you are available at that time, otherwise pleas discuss potential alternative dates in the wiki. So much from me for now, I will now begin my (hopefully deserved) one week winter holiday and hope to see you all in good health and humor in the next year. By the way: Merry Christmas to everyone who celebrates this occasion. Cheers, Sebastian P.S. I wont be available via Internet till Monday, December 30th. Please call me on my cell phone if you have important questions... --- [1] cvs -d :ext:[USER]@cleo-project.org:/var/lib/cvsd/repos co cleo_private

4 6

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Nsis_Imp January 2005