[Help-smalltalk] Re:

[Paolo Bonzini]

> Thanks for your replies.  The only disagreement we have so far would
> over the "vs." aspect of my decision.  From my perspective, it is a big
> decision with lots of lost work if I make a bad choice, and I must
> approach it like a business decision (which in fact it is).  Beyond
> that, sure, live and let write software.

Fair enough.

> Database connectivity: MySQL is the one to do :)  It would also be nice
> to have ODBC capability.

Yes, MySQL+ODBC (and possibly Postgres) are enough.

> SSL - that's ok, Squeak does not really have them either, though I
> suspect it will via the crypto group.  There is always stunnel too,
> though I prefer to have zero chance of configuration mistakes leading to
> use of clear text communications.  There is always OpenSSL and FFI.

Yes, though there are licensing problems and I'd prefer GNUTLS with FFI. 
   But we got the idea.

> Seaside?  That means you are doing continuations.

Sure I am.

>  That won't be the case, but I do need to ensure that I can: (1) create
> and use software for commercial purposes; (2) be able sell/license said
> software.  I have no problem with giving away improvements to the system
> itself

Fine.  If your software is not itself an improvement to the system, you 
can license it as you prefer.

> I strongly recommend finishing the graphical environment.  The essential
> components for the IDE are the browser, inspector, notifier (ok so far)
> and (ouch!!!) debugger.

As I said, this is a volunteer project.  Current volunteers are not 
focusing on the GUI, but they do not stop new volunteers to do so.  So, 
you can recommend finishing the GUI but you have to put your own energy 
into that. :-)

That said, it's not a huge task.  A surprisingly big part of the 
debugger is not necessarily tied to graphical operation.  There is a 
debugger -- it is slow and has also bitrotted with some probability, but 
it is there.  And there is a textual MiniDebugger, which shares some 
code (cut'n'paste) with the real thing.  It would be cool to refactor 
the common code into a debugging package.

> The latter is really the star of the show, and
> (forgive me) you do not have a complete Smalltalk system without it.

Agreed.

> The Dolphin approach is to
> force the user to define package boundaries; we get by just fine w/o a
> change sorter.

We do the same in GNU Smalltalk.  Package boundaries however are 
implicit in "what the files define".  You would need a tool to make 
boundaries explicit, i.e. to associate a class with the file it lives 
in.  What Dolphin does is probably a good model.

Paolo