[R-wiki] Summary of the discussion before creation of R-sig-wiki

[Philippe Grosjean]

Frank E Harrell Jr wrote:
> [...]


> Nice discussion Philippe, again.  Here are 3 points I'd like to make.
> 
> - I would not weight installation very much as you don't do it often and 
> users don't see this

I agree. It is just that I anticipate some people would like to install 
a local version of a part or whole of the R Wiki on their computer (for 
consultation of the documentation when not connected to the Internet). 
With DokuWiki, I did this by just unzipping \wiki somewhere in the \www 
directory. Even under Windows, it works with a plain installation of 
EasyPHP. I wonder if it can ever be that simple with TWiki.

> - Twiki stores only diffs so there is no explosion of disk space

??? What is the algorithm used for images? What is the CPU load to 
reconstruct pages from several diffs... or images from diffs?

I just check in DokuWiki. Everything is in \data.
Here is the content:

\data\pages are the latest wiki pages in clear (UTF-8 encoded plain 
text). The directory structure reflects the namespaces.

\data\media are images.

\data\attic stores successive versions of documents in gzipped formats. 
This is not diffs, it is really all versions of the documents.

\data\cache stores both xhtml cached versions of wiki pages (they are 
recalculated only if modified) and a special .i file that indexes words 
for rapid search.

\data\meta just stores [pages].indexed for all pages whose indexation is 
up-to-date.

Finally, \data\changes.log is a chronological list of page changes.
Personnally, I trust this because it is simple, and I can manage and 
hack this easily if needed. Also, it seems that performances are 
considered, i.e., page cache and indexation of words for fast search.

> - We base our whole department's web site on Twiki and it has been 
> extremely successful and free of software problems.  We have also had 
> great success adding plugins when we need them such as bibliographer 
> builders.

OK, so, I am convinced that it is a good Wiki engine. But I still wonder 
about performances for large Wikis. A priori, I tend to trust MediaWiki 
in this field, because it is the Wikipedia engine, and it runs smoothly 
with millions of Wiki pages!... But then, I read this: 
http://meta.wikimedia.org/wiki/DokuWiki_vs_MediaWiki_benchmarks. Gosh! 
This benchmark is from Wikimedia itself! So, when they say DokuWiki is 
faster than MediaWiki, I trust them!

Do you think we could set up Twiki and DokuWiki together on the same 
server for comparison with a few hundreds of pages (for instance, those 
we could get from .Rd files, once Rdconv will be modified to produce 
Wiki pages)? Could you make this test if I send you a complete DokuWiki 
install with (the simples) instructions to run it?

Best,

Philippe

P.S.: at the end of comparison between DokuWiki and TWiki at: 
http://wiki.splitbrain.org/wiki%3Acompare, you have:

"... But one of the “features” that I like so much in DokuWiki is the 
fact that is able to offer a wide/complete set of features in a small 
code footprint and looks nice (visually uncluttered). This, of course, 
is very subjective and is no technical advantage unless the code is done 
by following high quality practices." —  Straider 2005-02-24 12:15

and:

"I really don’t feel Twiki’s setup merits the rating of “Easy”"