Re: Licsensing scientific data
On Tue, Aug 23, 2011 at 7:09 AM, Andreas Tille <andreas@an3as.eu> wrote:
> On Tue, Aug 23, 2011 at 02:21:33AM +0100, Peter Cock wrote:
>> I have had an initial reply that the licensing issue will be raised
>> at the next UniProt consortium meeting, but in the meantime
>> they recommend downloading test files via the UniProt API
>> as part of a test suite, rather than bundling files.
>
> Seems to be a reasonable compromise specifically because you
> can be qute sure to fetch up to date files.
Agreed. Of course one of the downsides is you might want to
also test old data (the formatting of data files from public sources
like UniProt does change sometimes).
>> Obviously a unit test requiring network access has downsides,
>> but this seems a practical interim solution for packaging
>> EMBOSS etc within Debian free.
>
> Because we can not relay on network access when building the package we
> can not do this test at package build time. However, it is reasonable
> to assume that there is network access at build time (because in most
> practical cases you are fetching the package from network). So there
> is the option to ask a debconf question like
>
> Download test data from UniProt and run test suite after installation?
>
> [Y]es / No
>
There are likely to be other datasets that might need to be downloaded
at test time - I know EMBOSS has functionality to access records from
many many online resources, so I would think there are many possible
online tests that could be included. So how about a single more generic
prompt?
Should a similar prompt be added in the Biopython package script?
Our tests include some online tests by default (which should be
skipped gracefully if offline), but offline only mode can also be
requested (we use this in our nightly builds).
Peter
Reply to: