Python Popularity: Questions and Comments

[Dr. David J. Ritchie, Sr.]

Yes, someone pointed out in private e-mail that thanks to
the ability to supplant the import function one could implement
the ability to import from a remote node relatively easily.

The hierarchy for import of packages is the other way round compared to
http://my.node.com:port/path/to/module but I suppose that could
be worked out.

Also, as someone asked privately one would have to decided what
protocal would be used: ftp, http:, telnet: (just joking, I think). One could
even invent one's own protocol "pyt://" (still joking). I would
be inclined to think ftp because one is basically doing file transfers but
http: would have some interesting aspects such as the ability to
be importing a remote module into one's program and have
the import hyperlink to some other site and then to another and so on.

The security aspects would be pretty horrendous.  One would sort of
get driven to having the virtual machine within which this byte code
executes be in a sandbox and walled off from the
file system of the machine it is on and so
on.

Sounds like one is implementing java all over again to me (which
is of course how I got started--figuring out what features you would
have to add to Python to make it competitive (?) with Java) but it
is an interesting sand castle to build and then tear down when the
tide of reality comes in again...

Peter Hansen, P.Eng. of peter at engcorp.com wrote...

> The idea is sound, but for now I think really practical
>   only on a company intranet.  Internet connectivity is
>   still too unreliable to make this very effective
>   except for strictly Internet-related applications
>   (i.e. those which have no intrinsic purpose when an
>   Internet connection is not available.)
>
>   That said, I suppose a little local caching would provide
>   a reliable backup after the first invocation had succeeded.
>
>   And none of this would be in any way difficult to implement
>   even now, since you can just substitute your own implementation
>   of __import__().
>
>
I hear all the comments about security though and what would you do
if the machine you whose module you wanted was unavailable on the
net and so on.  These were the same things that were being thrown up
eight-to-ten years ago at the beginning of the web.  See my recent talk
about "Bringing the Web To Fermilab":

< http://home.fnal.gov/%7Eritchie/OTHERtasks/Bringing%20the%20Web%20to%20Fermilab.htm >

for some examples.  In particular, in preparing for the talk, one person told me
about the initial concerns thrown up by techies about the web being something
that no one would use because "what if the links weren't up?" and "what about the data
just being any old thing that someone put up and therefore wasn't valid?"

She said that of course all this about invalid and unreliable data was the case but in fact
the response of the average human being (as opposed to the techies)
was not to refuse to adopt the web but it was
instead something even scarier--they opted TO BELIEVE THE INFORMATION THEY
FOUND THERE.

So, with that example, the lack of reliability in a remote module import scheme may not
be something that kills it off.

Thanks for interesting discussions from all parties who sent me e-mail or responded
publically.

--D.

--
Dr. David J. Ritchie, Sr.
djrassoc01 at mindspring.com
http://home.mindspring.com/~djrassoc01/