Re: Website upload

[David Kastrup]

Urs Liska <address@hidden> writes:

> Am 07.03.2017 um 11:45 schrieb David Kastrup:
>> Urs Liska <address@hidden> writes:
>>
>>> Am 07.03.2017 um 11:24 schrieb Phil Holmes:
>>>
>>>> Again - this needs answering with a bit of care, because the website
>>>> and the docs are closely linked.  It appears to me that putting any of
>>>> the documentation (which includes part of the webite) is doen with
>>>> rsync without the delete option.  See
>>>> https://github.com/gperciva/gub/blob/master/test-lily/upload.py line 175. 
>>> OK, there's some general discussion in here.
>>>
>>> Basically I think such an upload should definitely include the
>>> --delete option because we don't want an arbitrary number of orphaned
>>> files on the server, isn't it?
>> We don't want orphaned files, no, but one better make a complete backup
>> first because not all files might be orphaned copies but may have been
>> added on the server only for some purpose.  Symlinks, referal files,
>> permission/login files and similar stuff.
>
> OK, that makes sense.
> But how would we find out about such items? I don't think it's a very
> clean strategy to make a backup, then remove them and deal with any bug
> report about missing files ...

The alternative is someone going through all the files by hand on a
backup/image/copy.  I think, however, that reviewing the top-level
directories should be enough: server-specific stuff is rarely found in
lower parts of the hierarchy only: so if we are missing something, this
is likely obvious in higher layers already.

-- 
David Kastrup