[Open-bio-l] Toolkits and the new eutils policies

[Peter]

On Thu, Mar 25, 2010 at 7:59 PM, Ewan Birney <birney at ebi.ac.uk> wrote:
>
> On Thu, 25 Mar 2010, Peter wrote:
>>
>> Thanks Chris (& Eric),
>>
>> Looks like we are fine as things stand: continue to
>> encourage the user to set the email (with a warning
>> if omitted), and try to encourage them to override the
>> tool parameter if appropriate (e.g. if part of a larger
>> application like a Galaxy workflow).
>>
>> [I don't see any point in forcing people to invent
>> tool names for each of their one off Entrez scripts,
>> or interactive sessions - defaulting to BioPerl etc
>> here seems sane]
>
> At the very least teh default should be "BioPerl Toolkit Placeholder For Non
> Registered Client" so that NCBI know precisely that the end programmer has
> not put something in there sensibly.

No, that's just silly IMHO. Using "BioPerl" on its own serves just
the same purpose (indeed, the NCBI will be used to this from
existing users and all versions of BioPerl to date), The extra
long version doesn't add any useful information and more
importantly makes the URLs much longer which can be a
real issue because long URLs can break (e.g. if going via
a proxy).

> And there should be a loud warning. I think it's fine to actually
> throw an exception. If someone is running a one off script, then they
> made the function call and can modify it. If someone's developing something
> more serious then they've got the time to think it through.
>
> I see little benefit in letting a default happen with just a warning.

Making the email and/or tool mandatory vs throwing an
exception just an implementation detail.

I think the issue is should BioPerl etc treat the email and
tool as optional, optional with a warning, or mandatory.
Note the NCBI does not seem to be making them
mandatory (for now).

Peter