[BioC] Re: offer from I Neuhaus to create a BMS response; keep in core, please

[Todd Richmond]

> >I can't imagine that a binary
> >format is going to make life easier for anyone. If they are
> concerned about
> >size, there are standard compression libraries that you can use to
> >compress/uncompress your data files on-the-fly as you save/load.
> >
> I disagree. A binary file could facilitate getting chunks of data using
> random access without having to go through the whole file. For example,
> you can imagine at some point creating you own 'probe sets' with
> features that are different than those mapped by Affy. However the
> structure of the file MUST be publicly available.

While that's certainly true, you can achieve the same goal with a plain text
file using memory-mapped files and/or indexing schemes. Of course, using
compression is probably out then. But disk space and memory are cheap - I'd
rather have readily accessible (and human-readable) text formats than
proprietary binary formats (even if their structure/format is publicly
available and well-documented).  Of course, I'm speaking from a purely
selfish point of view. We have our own equivalents of CEL/CDF files, though
ours are all just tab-delimited text files. If the affy package migrates to
binary files and proprietary APIs to keep in step with Affymetrix, it will
become more and more difficult for us to utilize the Bioconductor tools.
There's already a lot of affy specific assumptions in the oligo array tools
that cause problems for us - rightly so since it's the "affy" package. I'd
rather see a trend toward more generalized tools than tools that are highly
specific to a single platform. But that's because I'm lazy - I don't want to
have to re-code all of the "affy" package to make an equivalent "ngs"
package.

Regards, Todd

--
Todd Richmond
Manager of Research Informatics
NimbleGen Systems, Inc
One Science Court
Madison, WI 53711
Ph: 1-608-218-7651