etoys now available in Debian's non-free repository

Bert Freudenberg bert at freudenbergs.de
Sat Jun 21 06:41:52 EDT 2008


On 20.06.2008, at 16:36, Frank Ch. Eigler wrote:

> Holger Levsen <holger at layer-acht.org> writes:
>
>> [...]  The reason for having it in non-free is that the Debian
>> ftpmasters don't think it passes the criteria for inclusion in
>> main. [...]
>> http://paste.debian.net/6962
>> [...]
>
> (Sorry, this is probably OT for this list.)  Considering the age of
> this smalltalk-derived image, is there some reason to be convinced
> that it contains no code/data other than that could be regenerated
> from sources today? If there is a mechanical way of assuring that,
> how big a step would it be to re-create the image from those sources?


It could be done, but it's not like the bootstrap-and-snapshot as some  
Lisps are doing to reduce startup time. The images contains instances  
that were created interactively for which there is no source code. I'm  
not entirely sure how many of these instances there are, but it's not  
only a few for sure. It would be a huge undertaking, for no obvious  
benefit than to satisfy those who erroneously believe source code must  
come in text files. As one Smalltalker commented, "how quaint" ;)

There are efforts to rebuild Squeak from the bottom up, and we intend  
for Squeak 4.0 to be built on this:

	http://netjam.org/spoon/

But as you can see there, the goal is not to have source code in  
files, but to have a more modular system (the current image is quite  
monolithic). Indeed if you read the preface of the Spoon book linked  
above you'll see that the author set out to obviate even the text  
files that we still use to move code from one image to the other.

Anyway, the Debian ftpmasters did not even object to that, but they  
were concerned about how to be sure what changed from one image to the  
next. Squeak comes with all the necessary tools built into it, but  
this does not work well with their established work flow.

What could be done is to export all code from an image to a file tree,  
and then compare these trees. This could also be done with an external  
tool without running Squeak, for example, the PyPy guys wrote an image  
parser in about two days. But this would just generate diffs that  
could not be taken to produce one image from the other (it might work  
sometimes but not in general). Perhaps if there was this tool the  
Debian folks would be happy, but I'm not too optimistic.

- Bert -





More information about the Devel mailing list