[R-wiki] Gelman's comments about R tips vs Wiki

[Philippe Grosjean]

Hi,
Although I am ready to consider your point of view, I think I don't 
understand it very well.

1) Rtips is, by definition, a collection of short R code snippets. They 
are, of course, collected together in a logical way (i.e., sections with 
tips dealing with similar topics). Now, *why the hell would you like to 
have a linear presentation of these tips???* (is this what you mean by: 
"the nonlinearity of hyperlinks just adds confusion." ?) The best way to 
access these tips is by looking at keywords (the search box) and reading 
the tips related to your particular problem. So, why would you need a 
linear reading from tip A to tip Z? Who do read Rtips linearly from the 
first one to the last one? They are not made to be read like a book,... 
just because they are different and they serve a different purpose.

2) People critisize the fact that you have "disturing" items in the Wiki 
pages, that is mainly the top and bottom light blue toolbars with 
buttons like "Edit this page", "Old revisions", etc. These make part of 
a Wiki system, which allows you to edit and add material to these pages 
yourself. This is the main difference between a plain Web site and a 
Wiki site. Another difference is that, being the result of the work of 
several authors, a Wiki site tend often to be less organized than a Web 
site. We made a huge effort to get R Wiki well-organized, but always 
remember the differences between a Wiki site and a Web site.

3) The "one big page" model is not just hard for the server (indeed, all 
pages are cached by the Wiki engine), but they are harder to maintain in 
the Wiki approach: remember that all versions of Wiki pages are kept on 
the server, with revision history, etc. This mechanism is very efficient 
with lots of small pages (which is what a Wiki is designed for), but 
much less for one big page like: http://pj.freefaculty.org/R/Rtips.html. 
Everytime you change something in one tip, the version of the whole big 
page is archived. Also, when you track changes (revision history, RSS 
feed), you have warnings that "RTips was changed", no mather *which* tip 
was actually changed. So both RSS feed and revision history summary 
become rather inefficient when you tend to write large pages with many 
different topics in them. A Wiki is really designed to work with many 
pages, each dealing with a single particular topic. Now, you can flow 
linearly from one tip to the next one, if you like, by adding a "next 
tip" link at the end of each page.

4) I certainly hope that duplications (Web site versus same info in the 
Wiki) will be resolved. RTips has much to gain to be in a Wiki. One big 
reason is its maintenance. R is evolving quickly, and it is a huge work 
for authors/maintainers of R code AND R documents to keep them updated. 
Usually, R documentation (outside of the official R doc, I mean) is not 
updated regularly and is thus rapidly inconsitent with the current 
version of R. On the Wiki, any reader that founds a problem, an 
incompatibility, etc. can fix it. This brings a lot of help to the 
initial author(s) that did already a big job, but do not necessarily 
have the time to reread and retest all R code in their document for 
every new version of R (every 6 months).

So, yes, I hope that RTips on the Wiki will evolve with the output of 
other people and will become even more complete and up-to-date than it 
is in the original Web version. I would be in favor of complete close of 
the Web site version, with a redirection to the Wiki, but that is up to 
the Rtips author, of course. The fact is that the Wiki version would 
differ from the more static Web version (incompatibilites with new R 
versions and bug fixed, new tips added, etc.). So, why to point readers 
to an old, outdated version of a document? Point your readers to the 
latest version, that is, the Wiki one (well, once all RTips will be 
wikified)!

So, now, for people that do not like the presentation of the Wiki pages, 
I would suggest to build a web page with both a "linear" table of 
content (for instance in a frame at left and numbered tips as in the 
original RTips page), that points te the HTML version of Wiki pages (at 
right). To get HTML version of a Wiki page, you add '&do=export_xhtml' 
at the end of the URL. For instance, one RTip like:

http://wiki.r-project.org/rwiki/doku.php?id=tips:data-frames:recode_column

can be accessed as if it was a plain Web page using this URL:

http://wiki.r-project.org/rwiki/doku.php?id=tips:data-frames:recode_column&do=export_xhtml

Best,

Philippe Grosjean


Andrew Gelman wrote:
> dear ben,
> 
> as the author of two 600-page books i can appreciate the difficulties of 
> trying to be inclusive!  it's a paradox of the faq that as it becomes 
> more inclusive, it becomes harder to use.  thanks in any case for 
> organizing this!
> 
> yours
> andrew
> 
> Ben Bolker wrote:
> 
> 
>>  Andrew Gelman posted a link to "Rtips" on his web site.  Gregor
>>(Gorjanc???) pointed out that they were being moved to the Wiki,
>>and Andrew said he preferred the simpler page [see below].  This 
>>disturbs me ... I'd hate to maintain two different versions, and
>>I'm uncomfortable with the idea that we're putting the information
>>in a format where it is *less* accessible (at least to some readers). 
>>Is there any way we can tweak the design of the Wiki so more people like
>>Andrew will find it useful?
>>   Revisiting the discussion from a couple of days ago, where
>>Phillipe said that the "one big page" model was going to be too
>>hard on the server -- is there any way to make it easier?  Can
>>pages be cached?  Or should we all donate money toward upgrading?
>>(I do wonder a bit how well the server will manage if this
>>all takes off and it starts getting thousands of hits a day ...)
>>
>>  [in the mean time I have added an explicit link to R tips in the Wiki]
>>
>>  cheers
>>     Ben
>>
>>-------------------------
>>Comments
>>
>>Hello!
>>
>>These pages are now being moved to R wiki.
>>
>>Posted by: Gregor at April 17, 2006 11:53 PM
>>
>>Gregor,
>>
>>I find the R Wiki a bit overwhelming and actually prefer Johnson's 
>>simpler page. For me, there's a point at which the nonlinearity of 
>>hyperlinks just adds confusion. Perhaps this is one reason I still 
>>write books.
>>
>>I hope Johnson continues to maintain his webpage, or else I hope the 
>>Wiki people put in a prominent link to the old version.
>>
>>Posted by: Andrew [TypeKey Profile Page] at April 18, 2006 06:17 AM
> 
> 
>