Okay, guys, I presume Tom will clarify exactly what he did with extra
points, and I agree that with a specific forcefield it may be dangerous to
ignore something in the prmtop. However, and I hate to harp, but I feel
like no one is hearing what I am attempting to say, the more fundamental
problem is that there is potential ambiguity in the prmtop. This comes
about as we add more forcefields. There should be exact specification in
the prmtop as to what forcefield it applies to. Then, for each forcefield
there should be exact specification of what is needed in the prmtop for
that
forcefield, and what the data actually means (implying how it should be
used). Unless you put the effort into doing this, you will never get
things
right. Now, a little bit of testing is a good thing. BUT there is no
such
thing as complete testing; the permutations are mind-boggling and it is
not
at all possible. Correctness is essentially unprovable for anything but
the
simplist snippet of code. The only salvation is in clean design and
specification, especially of interfaces. This is actually not that hard
for
the prmtop. You go back and do the complete derivation for the '94 ff,
and
then everything is a delta from that base. BUT if you never do this and
keep adding forcefields, you are going to sink into a morass. This is
less
of an issue for me than it is for the amber group as a whole, because I
only
intend to implement forcefields that are really successful while in the
sand(er)box. BUT what is a killler for me is if any fundamental objects
in
the current prmtop are changed or removed. I think I may have
misunderstood
Yong, but the reason I jumped in on all this is that it sounded like that
was what he wanted to do.
Regards - Bob
----- Original Message -----
From: "Yong Duan" <duan.ucdavis.edu>
To: <amber-developers.scripps.edu>
Sent: Friday, November 11, 2005 10:42 PM
Subject: RE: amber-developers: Excluded list ?
>
> Carlos, thanks! Apparently, your English is better than mine :).
>
>
> yong
>
> -----Original Message-----
> From: owner-amber-developers.scripps.edu
> [mailto:owner-amber-developers.scripps.edu] On Behalf Of Carlos
Simmerling
> Sent: Friday, November 11, 2005 7:39 PM
> To: amber-developers.scripps.edu
> Subject: Re: amber-developers: Excluded list ?
>
>
> the key issue to me is not whether the list is supplemented
> but whether the content in the prmtop is ever discarded.
> I think that is what Yong initially suggested, and it's still not
> clear from these emails if part of the code is doing this or not.
> If someone doesn't want to use the prmtop list, they should
> be very careful in assuming that they know how it was built.
> For example, just because someone thinks that exclusions
> are for 1-2, 1-3 etc doesn't mean that you can just redo it
> that way. I think that this is a general concern- we really
> should not make any assumptions these days about what
> is "typical" or how something works without warning
> everyone else. Other developers are using the code in
> ways that are very different from a "traditional" MD
> simulation and assumptions about how the code runs or
> even what a "molecule" is can be dangerous. This
> doesn't have much to do with file formats, it's much
> more about assumptions about what the user's system
> must be.
>
> having said that, the best way to avoid problems is probably
> for us to write test cases that will break if someone makes
> incorrect assumptions. I'm not great about that myself,
> but it's probably the only practical way to protect our own areas
> of the code these days. dac has been saying that for a while now...
> carlos
>
>
>
Received on Wed Apr 05 2006 - 23:49:49 PDT
