Konstantin,

On Tue, Feb 26, 2019 at 4:23 PM Konstantin Kolinko <knst.koli...@gmail.com>
wrote:

> вт, 26 февр. 2019 г. в 21:24, Igal Sapir <i...@lucee.org>:
> >
> > Mark,
> >
> > On Tue, Feb 26, 2019 at 4:33 AM Mark Thomas <ma...@apache.org> wrote:
> >
> > > <snip/>
> > > branches
> > > - master (9.0.x)
> > > - 8.5.x
> > > - 7.0.x
> >
> > What are the plans for the tomcat-site?  [...]
>
> Igal,
>
> If you look at a *-fulldocs.tar.gz file with full documentation that
> accompanies each Tomcat release,
> you'll find that e.g. for Tomcat 9.0.16 its unpacked size is 60 Mb,
> with 2385 files in 165 directories.
>
> The full tomcat.apache.org site contains such documentation for every
> major version of Tomcat. See all directories listed ar
> http://svn.apache.org/viewvc/tomcat/site/trunk/docs/
>
>
With Subversion it is easy to work with that, using "sparse checkout"
> feature. The relevant commands are documented in site's README file.
>

Git is much more efficient than Subversion, as it only keeps a single copy
of files that were unchanged.  The new github repo takes on my drive 133.2
MiB for the 3 branches combined with the master branch checked out (125.6
MiB when the 7.0.x branch is checked out as less files get uncompressed).


>
> With some googling, it seems that Git also has some sparse checkout
> support [1], but I am not sure whether it is as easy to use as this
> feature in Subversion.
>
> (How does one manage the checkout layout? Is there a better way to
> update the ".git/info/sparse-checkout" file rather than having to edit
> it by manually? In Subversion the checkout depth can be changed easily
> with "svn update --set-depth" command).
>

In Git you don't have to get a partial depth due to the efficiency
mentioned above - you get the whole depth at no additional cost.  If you
have for example a file that hasn't been changed in any commit from 7.0.1
to 9.0.16 then Git keeps only one copy of it across all branches throughout
the repository.


> If people really want to use git for the site, I think the first step
> it to create a Git mirror for the current site repository and to see
> whether it is usable.
>

I have created a local git repo from the site branch.  From what I see the
most disk space goes for the JavaDoc sections, and some on the docs
directory. These are generated files so they can be added to .gitignore and
not tracked by Git.


> At this point, I think that Subversion is better tool for managing the
> site. I do not expect much from Git.
>

I agree with you that ideally the site should not be mixed with the source
code.  The tomcat-site should have its own repository.  A GitHub search
shows that the apache account has 94 repositories with `site` in the name
[1].  So for example there are repos like `kafka-site` and `spark-website`
in addition to the `kafka` and `spark` source code repositories.

All we have to do is create a separate repo for the site, either
`tomcat-website` or `tomcat-site`.  I believe that INFRA takes care of that.

Best,

Igal

[1] https://github.com/search?q=org%3Aapache+site&type=Repositories

Reply via email to