Philippe,
You are missing the point. So what if port 80 is difficult to use - use another one. The idea is to offer a system administrator an interface to check the status and make small changes to what is essentially an appliance. The fewer things that could potentially be mis-configured, the better. In the type of work I do, I will take simple, slow, and encrypted any day. I am not "building web sites" in this situation, so the rules to which you cling do not apply. That said, I will want your guidance for the few machines I envision setting up that will be "web sites" in the more typical sense. And yes, they will be based on Apache. "Apache will still be running when your image is long gone and can serve a nice 503 page". That's scary. Where is the problem? Is it Seaside, or dialect-specific? I am accustomed to my Smalltalk images running strong when the OS is starting to crumble around them. Bill ================================== Philippe Marschall philippe.marschall at gmail.com Sun Jul 13 06:59:56 UTC 2008 * Previous message: [Seaside] About Seaside 3.0 * Next message: [Seaside] About Seaside 3.0 * Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] You will have Apache as a frontend anyway because: - there is to my knowledge no way to run Squeak on a privileged port without giving it root rights - there is to my knowledge no way to generate Apache style access logs with Kom. The ones that can be read by a whole lot of tools like AWStats. - Apache will still be running when your image is long gone and can serve a nice 503 page - less bitching with your admin and security people Second, HTTPS/SSL really eats CPU cycles. Thats why a lot of sites like GMail use it only for login and then switch back to HTTP. Once you have accepted that there will always be an Apache frontend implementing a web server in Smalltalk doesn't make a whole lot of sense anymore. Something like an AJP connector will do fine. Cheers Philippe Wilhelm K. Schwab, Ph.D. University of Florida Department of Anesthesiology PO Box 100254 Gainesville, FL 32610-0254 Email: [hidden email] Tel: (352) 846-1285 FAX: (352) 392-7029 _______________________________________________ seaside mailing list [hidden email] http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside |
2008/7/13 Bill Schwab <[hidden email]>:
> Philippe, > > You are missing the point. So what if port 80 is difficult to use - use > another one. The idea is to offer a system administrator an interface > to check the status and make small changes to what is essentially an > appliance. The fewer things that could potentially be mis-configured, > the better. Nope sorry. HTTP is port 80 HTTPS is port 443. Anything else just causes trouble. > In the type of work I do, I will take simple, slow, and encrypted any > day. I am not "building web sites" in this situation, so the rules to > which you cling do not apply. That said, I will want your guidance for > the few machines I envision setting up that will be "web sites" in the > more typical sense. And yes, they will be based on Apache. > > > "Apache will still be running when your image is long gone and can serve > a nice 503 page". That's scary. No, that's very reassuring. > Where is the problem? Is it Seaside, > or dialect-specific? Smalltalk images do a lot of things as a consequence there are very lot of things that can go wrong. Smalltalk images were never engineered for reliability. One rouge thread can take down the whole image. Contrast that to Apache which was engineered for reliability. An Apache module can segmentation fault and Apache keeps running. Compare that to the Squeak image where the UI event thread can interact in very interesting ways with the GC. > I am accustomed to my Smalltalk images running > strong when the OS is starting to crumble around them. Good for you. Cheers Philippe _______________________________________________ seaside mailing list [hidden email] http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside |
I haven't been following the details of what Bill is proposing very
carefully but I can see the point he's trying to make. Many people may have internal tools written in Seaside. I have considered writing a web interface for user management at my current job. It would essentially be a tool for administrators and help desk staff to perform operations on accounts. Such an application would have a half dozen total users, maybe 2 concurrent at maximum. Given the sensitivity of some of the information, however, we would certainly want the connection encrypted. Such an application wouldn't need any of the power or reliability of Apache really and it certainly isn't a problem for it to be running on an alternate port (many of our existing administrative web interfaces--including vendor-suplied ones--run on alternate ports). As long as Apache is the easiest way to set this up, it's not the end of the world for me to do so. But if there was a solution available that didn't require the extra infrastructure I would certainly consider using it. Obviously this isn't going to be the best scaling architecture any time soon but that's ok. I also understand the desire not to have everything re-implemented (and then unmaintained) in Smalltalk. But if there's an itch that someone wants to scratch and enough users that it ends up being maintained, great. Julian _______________________________________________ seaside mailing list [hidden email] http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside |
In reply to this post by Philippe Marschall
I understand what you mean by reliability (running 20 servers and
killing the ones not running) but pay attention, squeak vms is not = to smalltalk vms. >> >> > > No, that's very reassuring. > >> Where is the problem? Is it Seaside, >> or dialect-specific? > > Smalltalk images do a lot of things as a consequence there are very > lot of things that can go wrong. Smalltalk images were never > engineered for reliability. One rouge thread can take down the whole > image. Contrast that to Apache which was engineered for reliability. > An Apache module can segmentation fault and Apache keeps running. > Compare that to the Squeak image where the UI event thread can > interact in very interesting ways with the GC. > >> I am accustomed to my Smalltalk images running >> strong when the OS is starting to crumble around them. > > Good for you. > > Cheers > Philippe > _______________________________________________ > seaside mailing list > [hidden email] > http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside > _______________________________________________ seaside mailing list [hidden email] http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside |
2008/7/13 stephane ducasse <[hidden email]>:
> I understand what you mean by reliability (running 20 servers and killing > the ones not running) No, I mean the following: For each request Seaside forks a special "restricted process" which has the following properties - it is not allowed to allocate more than 2 MB of heap memory - it is not allowed to use up more than 10 secs of CPU time (not counting IO wait and GC) - it is lightweight like normal Smalltalk processes In addation on a image level there are the following properties: - instances of WASession are not allowed to take up more than 100 MB total - we are not allowed to have more than 10'000 live instances of WASession - we are not allowed to have more than 20 concurrent Seaside requests -> we don't have more then 20 restricted processes Of course all numbers are subject to tweaking. > but pay attention, squeak vms is not = to smalltalk vms. Yes, gems are indeed very different and provide also other nice features like session replication. Cheers Philippe >>> >>> >> >> No, that's very reassuring. >> >>> Where is the problem? Is it Seaside, >>> or dialect-specific? >> >> Smalltalk images do a lot of things as a consequence there are very >> lot of things that can go wrong. Smalltalk images were never >> engineered for reliability. One rouge thread can take down the whole >> image. Contrast that to Apache which was engineered for reliability. >> An Apache module can segmentation fault and Apache keeps running. >> Compare that to the Squeak image where the UI event thread can >> interact in very interesting ways with the GC. >> >>> I am accustomed to my Smalltalk images running >>> strong when the OS is starting to crumble around them. >> >> Good for you. >> >> Cheers >> Philippe >> _______________________________________________ >> seaside mailing list >> [hidden email] >> http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside >> > > _______________________________________________ > seaside mailing list > [hidden email] > http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside > seaside mailing list [hidden email] http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside |
Philippe Marschall wrote:
> > No, I mean the following: > For each request Seaside forks a special "restricted process" which > has the following properties > - it is not allowed to allocate more than 2 MB of heap memory > - it is not allowed to use up more than 10 secs of CPU time (not > counting IO wait and GC) > - it is lightweight like normal Smalltalk processes > > In addation on a image level there are the following properties: > - instances of WASession are not allowed to take up more than 100 MB total > - we are not allowed to have more than 10'000 live instances of WASession > - we are not allowed to have more than 20 concurrent Seaside requests > -> we don't have more then 20 restricted processes > > Of course all numbers are subject to tweaking. > > > Excuse my ignorance, but is Squeak Seaside now doing all of the above? And if so, as of which version? Thanks! Nevin _______________________________________________ seaside mailing list [hidden email] http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside |
2008/7/14 Nevin Pratt <[hidden email]>:
> Philippe Marschall wrote: >> >> No, I mean the following: >> For each request Seaside forks a special "restricted process" which >> has the following properties >> - it is not allowed to allocate more than 2 MB of heap memory >> - it is not allowed to use up more than 10 secs of CPU time (not >> counting IO wait and GC) >> - it is lightweight like normal Smalltalk processes >> >> In addation on a image level there are the following properties: >> - instances of WASession are not allowed to take up more than 100 MB total >> - we are not allowed to have more than 10'000 live instances of WASession >> - we are not allowed to have more than 20 concurrent Seaside requests >> -> we don't have more then 20 restricted processes >> >> Of course all numbers are subject to tweaking. >> >> >> > > Excuse my ignorance, but is Squeak Seaside now doing all of the above? No, it's doing not one of them and that's the whole point. Cheers Philippe _______________________________________________ seaside mailing list [hidden email] http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside |
Free forum by Nabble | Edit this page |