Porting to VW

Michael S. Klein mklein at alumni.caltech.edu
Fri Nov 13 19:07:12 UTC 1998


> I'm new to the list so please forgive any reopening of old woundsbut has an
> XML-based
> source externalization been discussed?

Not here, really.  My personal take, having only given XML a cursory look,
is that XML has a rather large amount of overhead for Smalltalk source 
externalization.

1) The general overhead it has by trying to be 'human-readable' without 
really succeeding.  (I've never been able to read a bitstream without 
some device to help me).  I think people are deceived by the (relative) 
ease of externalizing strings.  If you ever have to deal with Kanji, things
are a lot harder, but still relativly easy. Personally, the advantage of 
being able to read source with an ordinary text editor just doesn't cut 
it with me.

2) XML seems to have a rather large amout of intellectual overhead to 
cater to the static-typing crowd.

On the upside: XML is a standard, and there is a fair body of (non-Smalltalk)
code for parsing/validation/...


Some problems with Squeak source:

The stamp identifies authors by their initials. Certainly easy, but but 
very robust.  ("This method was written by 'Dave Smith'"  :-)
I suppose that the community is small enough that name-space issues are still
pretty small, but someday...

The stamp mixes together the author and timestamp into a string

RunArray class >># scanFrom:

Why not just use the SmartRefStream to externalize source?

It would be nice if CompiledMethod didn't have slot-voodoo so that it
would be easier for us to experiment with alternate source/execution 
mechanisms. (Not to mention making VM allocation and GC cleaner)


-- Mike





More information about the Squeak-dev mailing list