WDSCI and EGL

From: Jon Paris

You can leave off the last part of the sentence Joe. It will never
generate RPG.

In reality there really is no reason for a generator to produce COBOL or
RPG
... Perhaps we should focus on getting themt o generate W-code just as the
RPG compiler does? It would make much more sense and since IBM have W-
code
oriented backends on all platforms, it should really be the cheapest
option
for them anyway. Why encourage them to generate something that can only
constrain EGL's capabilities to match those of the target language?

Well, as you know I'm still not sold on EGL as a business logic language.
If I squint real hard I guess I can see a direct EGL-to-W conversion; that's
not all bad and gets around some of the platform fanaticism. One of the
hard pieces would be linking the debugger with the high-level code.

Even if they got over that hurdle, though, I'm concerned that all I/O would
then be relegated to SQL. For example, even though EGL has nominal support
for indexed access, I've yet to figure out whether it would generate actual
indexed I/O for COBOL. Of course, this is a general concern, and really has
less to do with what they generate than with whether they have programmers
that understand the benefits of different database access methods.

But those concerns aside, I think I agree that EGL-to-W is the more correct
answer IF they decide that EGL is really going to be a language. I've had
some discussions with people on the team about what constitutes a "4GL" and
I keep pointing out that you can't do simple things like percolate database
changes to your JSPs. If the EGL team decides to promote EGL from a code
generator to a language, then they're going to need some better tools for
things like refactoring.

Even with java byte-code makes more sense than generating Java itself.

Yes, I suppose so, although having the source does make it easier to debug.
Of course, if the byte-code is properly structured a decent decompiler would
help a lot with that issue.


Joe