ikiwiki/doc/tips/spam_and_softwaresites.mdwn

[[!meta date="2010-03-01 13:14:48 +0000"]]

Any wiki with a form of web-editing enabled will have to deal with
spam. (See the [[plugins/blogspam]] plugin for one defensive tool you
can deploy).

If:

 * you are using ikiwiki to manage the website for a [[examples/softwaresite]]
 * you allow web-based commits, to let people correct documentation, or report
   bugs, etc.
 * the documentation is stored in the same revision control repository as your
   software

It is undesirable to have your software's VCS history tainted by spam and spam
clean-up commits. Here is one approach you can use to prevent this. This
example is for the [[git]] version control system, but the principles should
apply to others.

## Isolate web commits to a specific branch

Create a separate branch to contain web-originated edits (named `doc` in this
example):

    $ git checkout -b doc

Adjust your setup file accordingly:

    gitmaster_branch => 'doc',

## merging good web commits into the master branch

You will want to periodically merge legitimate web-based commits back into
your master branch. Ensure that there is no spam in the documentation
branch. If there is, see 'erase spam from the commit history', below, first.

Once you are confident it's clean:

    # ensure you are on the master branch
    $ git branch
      doc
    * master
    $ git merge --ff doc

## removing spam

### short term

In the short term, just revert the spammy commit.

If the spammy commit was the top-most:

    $ git revert HEAD

This will clean the spam out of the files, but it will leave both the spam
commit and the revert commit in the history.

### erase spam from the commit history

Git allows you to rewrite your commit history.  We will take advantage of this
to eradicate spam from the history of the doc branch.

This is a useful tool, but it is considered bad practise to rewrite the
history of public repositories. If your software's repository is public, you
should make it clear that the history of the `doc` branch in your repository
is unstable.

Once you have been spammed, use `git rebase` to remove the spam commits from
the history.  Assuming that your `doc` branch was split off from a branch
called `master`:

    # ensure you are on the doc branch
    $ git branch
    * doc
      master
    $ git rebase --interactive master

In your editor session, you will see a series of lines for each commit made to
the `doc` branch since it was branched from `master` (or since the last merge
back into `master`). Delete the lines corresponding to spammy commits, then
save and exit your editor.

Caveat: if there are no commits you want to keep (i.e. all the commits since
the last merge into master are either spam or spam reverts) then `git rebase`
will abort. Therefore, this approach only works if you have at least one
non-spam commit to the documentation since the last merge into `master`. For
this reason, it's best to wait until you have at least one
commit you want merged back into the main history before doing a rebase,
and until then, tackle spam with reverts.