- From: Christopher B Ferris <chrisfer@us.ibm.com>
- Date: Mon, 9 Jun 2003 06:14:44 -0400
- To: www-ws-arch@w3.org
www-ws-arch-request@w3.org wrote on 06/07/2003 11:53:54 AM: > > Glad to try. This may be particularly useful because getting to > specifics may surface misconceptions on my part. > > 1 - There is an application resident on a Web server somewhere that I > invoke, from an application on another server, with a simple GET with > some parameters. The service (or whatever you want to call it) does > something or other and returns a binary image (a png file) as an HTTP > image/png type. That's all it returns -- the image manufactured from > scratch using those parameters -- and that's all I need back. One of > the two things I had difficulty with in getting this working (the other > is not relevant to this discussion and SOAP certainly would not have > helped that problem) was figuring out how to handle the binary stream of > data. Eventually I discovered some method in the system library that > would just stream it out to a file on disk, which is what I wanted. I > believe that my task would have been significantly more difficult if I > had also had to cope with some XML SOAP header that came along with the Only if you tried to handle that level on your own. It would be no different than trying to implement the HTTP protocol stack on your own. But this is the realm of middleware. Most SOAP toolkits bypass all this gorp and give the application programmer something that is far more natural for them to use. > image. That SOAP header would not have done me any good whatsoever, and > I am not interested in any of the addons that it might offer (like > signing the image, for God's sake). Streaming the header plus the > binary image to a file would definitely not have been what I wanted. Right, so don't use it, but again, you are implying that most developers would be doing all this gorp on their own which is not likely under most circumstances. > > 2 - We have an automated XML feed from our Help desk to a service > company. I'm not positive how it works, but I think it just does a > simple POST and slams the XML into something that stores it as-is. For > one reason or another that's ALL that is needed. SOAP headers would > simply be added characters that would be thrown away, and presumably > software would have to be written to strip those headers off. Fine, then don't use SOAP. However, in the event that what you really wanted was a direct feed from the source to whatever it is that reads periodically from the filesystem the XML that was simply dropped into the file system as a result of the POST, then IMO the types of APIs generated by most Web services toolkits would be far more natural for the developers. > > In both cases, note that the result of the "push" or the "pull" is > simply to take a defined set of data and put it somewhere. The data are > not processed at that time. SOAP stuff would require processing to > strip it off. Cost with no benefit. Well, like I said, if there was a requirement to eliminate some of the time from the processing of the complete application (e.g. eliminate the time where the information simply sits idle on the filesystem), then you might want to reconsider this. Additionally, if this is indeed just a pitstop, then for security reasons, you might not want to have the information simply sitting on the HTTP server's filesystem. It is more prone to risk of attack. If it were me, I might have a direct feed into my pub/sub or message queue infrastructure, such that the messages would trigger the down stream processing and such that the information were not sitting on my HTTP server's filesystem waiting for some hacker to find and exploit in ways that I might not appreciate. > > It sounds like Hao can probably come up with better scenarios, and I > hope he does so. > <snip/> Cheers, Christopher Ferris STSM, Emerging e-business Industry Architecture email: chrisfer@us.ibm.com phone: +1 508 234 3624
Received on Monday, 9 June 2003 06:14:57 UTC