>I have a problem that I haven't been able to solve, and I thought--again, >since I've posted a brief version of a part of this issue before--that >perhaps someone on the list might be able to help. And sorry I had to make >it so long-winded.... > >Problem: I have a dual server configuration where a dedicated "slave" server >(500 Mhz G4 Mac) is giving database services by being directly connected to >a database-application/internet server (NT dual 700 Mhz Dell) listening for >WAN/http calls for data. MC is a custom browser asking for db record data. >While in a true WAN environment, approximately 5-10% of my MC-generated >"post" calls never generate a return from the http server, even though the >database records being called for are clearly found and delivered to the NT >(as apparent in the db application's "found records" status when looking at >the NT server directly). Netscape/IE _never_ fail to generate returns when >we query the db through the web, which throws some suspicion on MC. > >Details are: > >I am using post commands similar to one provided below to communicate >through lasso (CGI application from BlueWorld) with FileMaker Pro >application/database located on the web/db servers. FileMaker Pro database >is used as a back end for my MetaCard stacks on the client machines. > >I moved this configuration to a relatively quite LAN with very good >bandwidth for benchwork evaluations and the failures still persist, only at >about a .5-1% rate instead of the 5-10% rate. I am still evaluating setups >and all that jazz on the NT to see if the NT is the problem, but still have >to wonder whether MC is in any way at fault or can be used to generate a >"work-around" solution. For example, if MC just didn't "hang" when the >response doesn't come back in, I could check for returned results, if >finding empty-- re-send the query and go on. > >OK. > >I did a little experimenting and it appears that instead of using the "post" >command I may use a "put" command as well. > >This means that instead of > > post >"-database=Students.fp3&-layout=FirstStep&-response=Field:'gbRespOne'&-recor >dI >D=48&-search" to url "http://##. ##. ##. #/FMP_MC/Action.lasso" (of course a >real IP nbr goes where ##'s are) > >I can simply use > > put url >"http://##. ##. ##. >#/FMP_MC/action.lasso?-database=Students.fp3&-layout=Firs >tStep&-response=Field:'gbRespOne'&-recordID=48&-search" into fld "report" > >with the same result. Now, I haven't had the luxury yet of testing whether >this improves on the percentage of unreturned calls, but the following issue >comes up regardless: > >If I recall various sources of information correctly, Scott has suggested >that MC is likely to lose the "post" command from the next >version of MC in favor of socket-level coding. > >Scott -- is this correct? If so, is the "put url" also likely to go? Is >there a difference in how they work underneath? > >But more importantly..... >because I only partly understand the mechanics of using "sockets-level" >programming for http, as opposed to post/put commands, I would like to offer >the following question to the list in hopes that you all can give me some >more clues about "sockets." > >Q1) If it is the case that "post" is untrustworthy or about to become >extinct, then what will be the "socket" equivalent of the >following post command? It's really not that scary. The higher level of control/monitoring over what is going on makes it worth learning. In your example you would do something like this: open socket <xxx.xxx.xxx.xxx:80> ##open connection to the server on port 80 write <stuff> to socket <xxx.xxx.xxx.xxx:80> ##write the POST stuff just like a browser would read from socket <xxx.xxx.xxx.xxx:80> until eof ## get the result from the server. put it into var1 A while back I answered a similar post with *real* examples, look through the archives for this month or the one before. I would also suspect the other side - server+lasso. I would also try POSTing the whole header just like a browser would. There is a mac tool OTSessioWatcher by Stairways which allows you to follow what exactly is sent and received in a network communication which can show you what a browser sends and what the server replies. > >post >"-Database=Students.fp3&-layout=LayoutNameHere&-Response=ResponseFileNameHer >e& >Social_Security=333444333&-search" to url "http://##. ##. ##. >#/Action.lasso" > >This Post command will search for #333444333 in Social Security field in >Students database and, through a response file, return the student's first >name, >last name etc. > >>From a browser's perspective the HTML equivalent of post command above (in >the form of the web link) will be: > > <A >HREF="action.lasso?-database=Students.fp3&-layout=Details&-response=Studen >tInfo.lasso&Social_Security=333444333&-search">Click this link to retrieve >students Information</A> > >So, the number 1 Qustion really is what will be the "socket" equivalent of >the following >post command? > >post >"-Database=Students.fp3&-layout=LayoutNameHere&-Response=ResponseFileNameHer >e& >Social_Security=333444333&-search" to url "http://##. ##. ##. >#/Action.lasso" > >Q2 through 5) Why are socket-level solutions, when finally written, likely >to be more reliable? Are they truly more flexible? Why is the required >"programmer's" knowledge load being moved from "built-in" to >"build-your-own" with these improvements? Will this level offer, for >example, one to trap a failed return so a call can be reissued? This is a >complex mix of MC rules and http rules that I'm trying to get a handle on >and somehow haven't yet fully grasped. And I am willing to try to learn it >if it really will solve my problem. Where can I start? I suspect we have a >few really savvy internet MC developers out there who can help, if not by >explanations at least by offering working examples, and any help would be >greatly appreciated, either on or off list. > >Thanks, >Roger > Regards, Andu __________________ [EMAIL PROTECTED] Archives: http://www.mail-archive.com/metacard%40lists.best.com/ Info: http://www.xworlds.com/metacard/mailinglist.htm Please send bug reports to <[EMAIL PROTECTED]>, not this list.
