Understanding alias (used as rewrite)

[Francis Daly]

On Wed, Jun 24, 2015 at 05:19:52PM -0700, E.B. wrote:

Hi there,

> > Could you explain?
> 
> Sure I wanted to take a prefix:
> 
> /my-long-base-path
> 
> and make sure all uris that match
> that prefix are served to the same
> file.

For this mail thread, that's probably fine. But for a general solution,
you'll want to make sure that you are clear about what "served to the
same file" means.

If it is "serve a file from filesystem", then alias and/or try_files may
be useful. If it is "send to an upstream such as via fastcgi_pass",
then alias may not be needed at all.

> I got a rewrite to do that perfectly
> but I see a lot of talk that rewrites
> are what you used in apache and are
> slow and not the nginx way of thinking.
> and it make sense that because the
> target file is fixed that avoiding
> the rewrite engine will be faster just
> pointing to a static aliased file.
> 
> I digress but hearing your opinion on
> this is interesting to me.

Generally, if I can avoid rewrite but achieve the same result simply,
that's what I'd tend to do.

But without knowing what the original rewrite was doing, it was hard to guess.

> > "alias" is, as you noted, documented at http://nginx.org/r/alias. It is
> > used to identify a filename that should be used to handle the request.
> 
> Im still wondering why the critical part
> you kindly explaining below is not documented.
> Isn't this basic feature and important info
> to know?

My suspicion is that what is documented is clear enough to the writer;
and no reader had previously pointed out any problems.

(I think that my explanation does follow from the documented words;
but I can see how it may not be immediately obvious. So if the
nginx documentation team is reading: this thread has some suggested
alternate/additional words for the documentation of "alias".)

> i discovered in the fastcgi conf file there
> was a:
> try_files $uri =404
> 
> so it overrided the alias and causeding
> the problem.

Yes; "alias" and "try_files" may not work together as you may immediately
expect.

It doesn't actually *override* the alias; but it does do (and fail)
a test that you probably do not want it to do.

(And that try_files line is part of "the configuration that is not yet
shown". The "include" file could have almost anything in it which would
change the way that the configuration is understood.)

> i guess this was a security
> measure to prevent sneaking around the
> filesystem for php requests.

I confess I've never been quite sure of the point of that line.

I can see what it does, and I think that it might be useful in some
limited circumstances which include "...and my php is configured badly
and I won't change it..."; but I've tried to avoid those circumstances.

> is there a
> better way to effect same protection?

If you can specify what you consider the "same protection" to be,
then maybe.

And kudos for correct use of the verb "to effect" ;-)

> try_files $request_filename =404????

That won't do what you want because of how try_files handles its not-last
arguments.

Possibly in this one specific case -- so not in fastcgi.conf that is
included elsewhere --

  try_files "" =404;

would do it. But you know that you are sending SCRIPT_FILENAME (or
whatever your fastcgi server honours) set to one specific filename only,
and you know that the matching file exists. So what is the test doing
that would be bad if it were not done?

Cheers,

	f
-- 
Francis Daly        francis at daoine.org