Re: Trial for RDi SOA

Joe Pluta skrev den 15-05-2008 22:44:

Interesting. In my experience, an essential part of any project is creating the database. In fact, design issues dictated several features of the database layout (especially the logging!). Heck, the SQL people say you can just add columns and such on the fly, it's no problem! :)

Great. So all this is not a matter of having similar implementations on (almost) the same data for comparison, but reinventing the wheel for show.

A hard problem is actually database refactoring which _should_ be as easy as code refactoring, but isn't. Then the SQL people would be right.

Personally I do not say that EGL is better or worse than anything else, I just say that _I_ am very careful of meta-languages since the additional translation step may be extremely hard to debug. Just look at how long it took before JSP's were debuggable - that was painful.

EGL is debuggable both at the EGL level and the generated code level (Java or JavaScript). It's really quite powerful.

I am not doubting it is very powerful. I am trying to say that in case of problems it might be hard to debug backwards to the original EGL from a user report.

Honestly Thorbjørn I would like to see ANY of the technology advocates take a stab at this. As I said, I'm truly tired of hearing people talk about how wonderful things are (and how awul Jav/EGL/IBM are). I'd like to see somebody prove it.

Well, it isn't exactly rocket science (seen from the web world) but still requiring a lot of work. I do not have a toolchest ready for AJAX web applications so I am not the right one to ask.

My guess is that the ONLY folks who might make a decent stab at this are the Rails folks. They have a pretty good framework, although once you step outside the framework, it tends to choke. I'm pretty confident that there won't be a PHP entry in this, nor will there be an RPG-CGI equivalent. It's just too hard to do both sides of the application in either technology.

The .NET folks might do it as well, but every action would require SQL to the System i database, and that would be horridly slow.

I have seen MySQL easily handle 1200 simultaneous users on a 4 CPU machine spewing out SQL requests. There is nothing intrinsic in SQL meaning that it NEEDS to be slow - as always it depends on how you use the tool. SQL on the i might be slow, I have not worked with it much yet.


I might take up the challenge, but I don't think so. If I get to do such a webapp at work, I'd be happy to share the results. My stuff would be in Java and run on the i.