R help - Jun 2009 - Software lifecycle for R releases (aka practical limits of support for older versions)

Hello useRs:

Does anyone have thoughts on the lifecycle of older releases of R? I
know that currently the 2.8.x and 2.9.x releases seem to be actively
"supported" on the mailing lists, but what about older releases, say
2.4.x? Curious to hear when people think older versions of R become
obsolete and unsupportable on the lists (or other venues). 

Regards,
Brian


--------------------------------------------------------------------------
This message w/attachments (message) may be privileged, ...{{dropped:27}}

There is an archive for all packages for older versions of R, but if
you want up-to-date  functionality of packages then you need the
newest versions.
my 2 cents

stephen

On Thu, Jun 25, 2009 at 12:38 PM, Rowe, Brian Lee Yung (Portfolio
Analytics)<B_Rowe at ml.com> wrote:
> Hello useRs:
>
> Does anyone have thoughts on the lifecycle of older releases of R? I
> know that currently the 2.8.x and 2.9.x releases seem to be actively
> "supported" on the mailing lists, but what about older releases,
say
> 2.4.x? Curious to hear when people think older versions of R become
> obsolete and unsupportable on the lists (or other venues).
>
> Regards,
> Brian
>
>
> --------------------------------------------------------------------------
> This message w/attachments (message) may be privileged...{{dropped:21}}

On Thu, 25 Jun 2009, Rowe, Brian Lee Yung (Portfolio Analytics) wrote:
> Does anyone have thoughts on the lifecycle of older releases of R? I
> know that currently the 2.8.x and 2.9.x releases seem to be actively
> "supported" on the mailing lists, but what about older releases,
say
> 2.4.x? Curious to hear when people think older versions of R become
> obsolete and unsupportable on the lists (or other venues).
Opinions vary, but:

- reporting bugs (or asking if something is a bug) based on any older version of
R than 2.9.x would likely get you flamed.

- if your problem could be solved by updating to the current version, I think
you would be expected to do so.

My personal feeling is that you can just about get away with updating R only
annually. Since you can easily keep an archive of previous versions available,
there's no need to avoid updating on that account.

Based just on R itself a longer update delay might be ok, but CRAN doesn't
supply binaries of new or updated packages for old versions of R. Many packages
will become seriously outdated much faster than base R.

     -thomas

Thomas Lumley			Assoc. Professor, Biostatistics
tlumley at u.washington.edu	University of Washington, Seattle

On Thu, Jun 25, 2009 at 9:38 AM, Rowe, Brian Lee Yung (Portfolio
Analytics)<B_Rowe at ml.com> wrote:
> Hello useRs:
>
> Does anyone have thoughts on the lifecycle of older releases of R? I
> know that currently the 2.8.x and 2.9.x releases seem to be actively
> "supported" on the mailing lists, but what about older releases,
say
> 2.4.x? Curious to hear when people think older versions of R become
> obsolete and unsupportable on the lists (or other venues).
>
> Regards,
> Brian
This is actually a fairly common question from R users at commercial
institutions, where for various reasons upgrading to the latest
release of R isn't always possible. This might be for regulatory
reasons (only a certain distribution of R has been validated), because
of IT policies, of because R is incorporated into a production
application where the risk of breaking the application outweighs the
potential benefits of an upgrade.

This is actually one of the main reasons why the release cycle for
REvolution R isn't as frequent as that for CRAN R. The beneficial side
effect is that we can therefore provide support for older versions of
R in our distributions. We support R 2.7.2 through our distribution of
REvolution R Enterprise, for example. More info here:
http://www.revolution-computing.com/products/revolution-enterprise.php

# David Smith

-- 
David M Smith <david at revolution-computing.com>
Director of Community, REvolution Computing www.revolution-computing.com
Tel: +1 (206) 577-4778 x3203 (San Francisco, USA)

Check out our upcoming events schedule at www.revolution-computing.com/events

On Jun 25, 2009, at 11:38 AM, Rowe, Brian Lee Yung (Portfolio  
Analytics) wrote:
> Hello useRs:
>
> Does anyone have thoughts on the lifecycle of older releases of R? I
> know that currently the 2.8.x and 2.9.x releases seem to be actively
> "supported" on the mailing lists, but what about older releases,
say
> 2.4.x? Curious to hear when people think older versions of R become
> obsolete and unsupportable on the lists (or other venues).
>
> Regards,
> Brian
For a description of R's formal SDLC, read:

   http://www.r-project.org/doc/R-FDA.pdf

While that document is targeted to R users in the domain of regulated  
clinical trials, much of the content is relevant to other general use  
domains.

 From the perspective of getting community support on R-Help, if you  
are using version 2.4.0 and you post a version independent query to  
the list, you will get a helpful reply, especially if you don't  
include in the post that you are running version 2.4.0.

However, it is possible that in the replies, there may be references  
to functions, function arguments or packages that are part of or are  
designed for newer versions of R. Upon reading that reply, you may end  
up scratching your head, wondering why you cannot find them in your  
version, which may prompt you to reply requesting clarification. That  
may lead you down the path to the next scenario...

If you include in your post (or a follow up) that you are actually  
using version 2.4.0, you will get a series of rather curt  
recommendations to update to the current release version of R included  
in any responses to your query.

However, if you post a query pertaining to what you perceive as a bug  
in 2.4.0 or a more recent version (possibly even 2.9.0 with 2.9.1  
imminent), you will get a pretty rapid stream of replies, with a level  
of hostility (flaming) included. Those replies will tell you in no  
uncertain terms, that you better upgrade to the most recent version of  
R (which may include a "patched" version) before reporting bugs  
against versions that from a development standpoint, are no longer  
supported. You would be expected to check the most recent version that  
you can install to see if the behavior that you perceive as a bug is  
still present.

The worst case scenario perhaps, in terms of being on the receiving  
end of flames, would be to actually submit a formal bug report on an  
older version of R, as that requires a **volunteer** member of R Core  
to have to stop what they are doing and spend time manually  
administering that report.

Finally, a good reference to go along with this general discussion, is  
the Posting Guide, listed at the bottom of all e-mails coming from the  
list:

   http://www.r-project.org/posting-guide.html

HTH,

Marc Schwartz