On Tue, Mar 24, 2015 at 7:14 AM, Martin Morgan
<mtmorgan at fredhutch.org>
wrote:
On 03/24/2015 02:31 AM, Wolfgang Huber wrote:
Before we start a religious war, can we make progress on the
pragmatic
goal of making it possible to provide such URLs to people?
There are two concepts
- ?the package' - a specific version, running in a specific
environment,
?frozen?, etc. (Gabe)
- ?the package? - as a concept and a living artifact (me, Bernd,
Tim)
Both are useful. And having URLs for both would also be useful.
0. That's (mostly) satisfied with the current scheme and
http://bioconductor.org/packages/3.0/bioc/html/BiocGenerics.html
http://bioconductor.org/packages/release/bioc/html/BiocGenerics.html
http://bioconductor.org/packages/devel/bioc/html/BiocGenerics.html
(hey, no www. -- that's four letters already! Perhaps importantly,
there's
also a hard-coded version for devel, 3.1, and for past releases. So
as I
understand it the request is for (a) shorter path names and (b)
dynamic
selection of release vs. devel, mentioned below, for the <6 month
period
when the package is in devel but not yet release. Also noted is
Henrik's
earlier proposal mentioned by Sean.
1. 'packages', 'bioc', 'html' all look somehow redundant, so
http://bioconductor.org/release/BiocGenerics.html
http://bioconductor.org/devel/BiocGenerics.html
http://bioconductor.org/3.0/BiocGenerics.html
but also
http://bioconductor.org/release/BiocGenerics/ (implicit
index.html)
http://bioconductor.org/BiocGenerics/release/
and their devel and version counterparts would seem quite possible
/ not
profoundly controversial. Landing pages for specific versions
3.22.7 do
not currently exist, change little across package minor versions,
and would
not lead to packages installable via biocLite(), so this idea of
Tim's is a
non-starter in my opinion.
Having the 'version' level of the path before the package provides
a
logical place to put biocViews for that release. I'd vote for one
of the
release/BiocGenerics[.html] schemes.
2. Something like
http://bioconductor.org/BiocGenerics
redirecting to release when available, devel when newly added
(Wolfgang's
proposal) would in my opinion be confusing, especially since we
continue to
have so much difficulty with version mismatches in user
installations. I
don't think having a warning on redirect that 'this package is not
available for release' would be effective either at advertising
robust
software or at enabling use by comparatively naive users.
3. In terms of the 'redundant' parts of the path, these are not
completely
arbitrary (not that these considerations have to dictate
presentation; they
do make one suspect that 'add a redirect and everything will be
fine' will
result in a nice plate of spaghetti, especially if there is some
desire to
retain backward compatibility).
'packages' separates the package repository from other aspects of
bioconductor.org, and group related concepts ('package', 'help',
etc.) at
a similar hierarchical level.
'bioc' serves to distinguish between software ('bioc/'), annotation
('data/annotation') and experiment data ('data/experiment')
packages, and
these divide the overall repository into three for the purposes of
biocLite() / install.packages() (this conceptual distinction has
been
useful, I think).
BioCsoft
"http://bioconductor.org/packages/3.1/bioc"
BioCann
"http://bioconductor.org/packages/3.1/data/annotation"
BioCexp
"http://bioconductor.org/packages/3.1/data/experiment"
'html' distinguishes the landing pages from the package tar balls /
binary
distributions themselves as returned by
contrib.url(biocinstallRepos()),
and from their vignette/, man/ and news/ resources.
4. In terms of best practices, it seems like articles are about
particular
versions and should cite the package as such, for instance if only
in devel
when the paper is being written as .../3.1/..., but that there is
no
substantive cost to also referencing 'current version available
[after
April, 2015] at .../release/....
5. At the end of the day I find myself casting my lot for landing
pages
with the form
http://bioconductor.org/release/BiocGenerics/
which leads to a little less typing but not the dynamic resolution
that
started this (version) of the thread.
Martin
Wolfgang
On Mar 23, 2015, at 18:43 GMT+1, Tim Triche, Jr.
<tim.triche at gmail.com>
On March 23, 2015 9:18:57 AM PDT, "Tim Triche, Jr." <
tim.triche at gmail.com>
wrote:
Packages are (read: should be, IMHO) published, citable pieces
of
research, though. Imagine if a paper you cite were silently
updated
without the doi/citation changing. That wouldn't be good
I don't disagree, but the existing setup does nothing to
address that.
Citation('limma'), for example, does.
.../release/... and .../devel/... can change at any time,
potentially
overnight (with or without a new BioC release). The only real
way to
cite an exact version is to cite that exact version, which is
already
the proper way to do things and would remain unaffected by
this, at
least AFAIK.
Perhaps a useful addendum would be for the mnemonic
http://bioconductor.org/limma
To redirect to
And then everything is explicit.
Does that address the competing issues discussed herein?
Note that 'release' and 'devel' are just symlinks to the current
release
and devel versions. I.e. currently 3.0 and 3.1 respectively. So
you can
always link directly to a specific version.
Dan