[Pipet Devel] Fwd: Bioperl: Pise release 5.a

[Brad Chapman]

Sorry so slow to respond--I just finished exams this week and had a 
chance to check Pise out more.

> Pise looks like a good model for us to study when we imeplement
> our own web-based interface to Loci.

I don't know, I wasn't that excited by the web user interface (I 
scoped it out with the url that Jennifer send). The thing that I 
didn't really like is that:

1. It really hides the command line from the user, so that even if you 
used Pise for a while, you would still never really have an idea how 
to run the same program on the command line. I think a nice thing 
about Loci is that it is kind of a way to teach a user how to work 
with the command line. 

2. The help information for a program is separated into a big chunk at 
the bottom of the page. If you take a look, for instance, at the 
advanced FASTA page, there is a whole bunch of info at the bottom, 
which seems just about as fun to look at as a man page.

It supposedly has x windows/Tcl interfaces as well, but I didn't see 
any screen shots of them (and didn't fire Pise up on my computer). 
They also mention a "coversational" interface, which I assume is like 
the nli interface that Gary and Jeff are proposing, but didn't see any 
more about that either.

> From my cursory rip through
> throught the description, it differs from loci primarily in that the
> data redirection (which is not well defined, so I could be off here),
> seems simplistic.  From what I can tell it is not 
network-distrituted,
> simply piped.  

Right, it seems like they are saying that it is just a simple copy, 
and not even really piped (just analagous to pipes). 

> This implies that the processing applications reside on
> the same machine. The piping appears step-wise.  Consequently the
> communications layer is not abstracted, and thus limited to 
command-line
> programs.  Loci's strength is in its generalized data-connectivity
> brokering capability, 

Of course, implementing that will be not so easy :-0 But it will be 
cool :) It seems like Pise is just meant more for users 
afraid/unwilling to use the command line, and wouldn't be used by more 
advanced users. Hopefully, I think Loci will have enough power that 
advanced users will see an advantage to using it (much as many 
advanced users see an advantage to using an X windows system over 
starting at black and white all day).

> so I don't see Pise as a 'threat' so much as a
> 'resource': Pise is GPL, so we are free to use it and study it, and  
 It
> does appear to have some very good design points (similar to applab 
in
> many respects).

They use XML descriptions of programs to generate the output, which I 
think are really worth taking a look at. They've already got a lot of 
programs running with this, and it might be worthwhile to just use 
their XML to generate our own interfaces (plus better piping), 
provided no one finds their XML system too distasteful. I think this 
would be a good place to start for this, and then we can think about 
how to make Loci a development environment so that you can make these 
XML documents in the context of Loci.

    Of course, the code is hundreds of lines of undocumented perl. 
Fuuuuuuun reading! 

Brad