You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by Eric Covener <co...@gmail.com> on 2008/07/21 14:09:23 UTC

Re: svn commit: r678179 - /httpd/httpd/branches/2.2.x/STATUS

On Sat, Jul 19, 2008 at 12:03 PM,  <ni...@apache.org> wrote:

>  * mod_auth_digest: Detect during startup when AuthDigestProvider
>    is configured to use an incompatible provider via AuthnProviderAlias.
> @@ -164,6 +149,10 @@
>    Backport version for 2.2.x of patch:
>         http://people.apache.org/~covener/2.2.x-digest_vs_authnalias-2.diff
>    +1: covener, rpluem
> +   -1: niq: looks like a major MMN bump to me, in that it'll
> +            break binary compatibility with old compilations
> +            of authn modules (has_realm_hash being undefined).
> +            Why not just test get_realm_hash - as in the else clause?

Is it kosher to pull the API extension from trunk? A less intrusive
(AuthnProviderAlias-only) solution occurred to me while reading your
note.  Would the (incorrect) minor bump be left in place or can it be
reverted?

(AuthnProviderAlias can just register a different authn_provider
struct and not put a reference to its own own get_realm_hash() in it)



-- 
Eric Covener
covener@gmail.com

Re: svn commit: r678179 - /httpd/httpd/branches/2.2.x/STATUS

Posted by "Plüm, Rüdiger, VF-Group" <ru...@vodafone.com>.
 

> -----Ursprüngliche Nachricht-----
> Von: Eric Covener 
> Gesendet: Montag, 21. Juli 2008 14:09
> An: dev@httpd.apache.org
> Betreff: Re: svn commit: r678179 - /httpd/httpd/branches/2.2.x/STATUS
> 
> On Sat, Jul 19, 2008 at 12:03 PM,  <ni...@apache.org> wrote:
> 
> >  * mod_auth_digest: Detect during startup when AuthDigestProvider
> >    is configured to use an incompatible provider via 
> AuthnProviderAlias.
> > @@ -164,6 +149,10 @@
> >    Backport version for 2.2.x of patch:
> >         
> http://people.apache.org/~covener/2.2.x-digest_vs_authnalias-2.diff
> >    +1: covener, rpluem
> > +   -1: niq: looks like a major MMN bump to me, in that it'll
> > +            break binary compatibility with old compilations
> > +            of authn modules (has_realm_hash being undefined).
> > +            Why not just test get_realm_hash - as in the 
> else clause?
> 
> Is it kosher to pull the API extension from trunk? A less intrusive
> (AuthnProviderAlias-only) solution occurred to me while reading your
> note.  Would the (incorrect) minor bump be left in place or can it be
> reverted?
> 
> (AuthnProviderAlias can just register a different authn_provider
> struct and not put a reference to its own own get_realm_hash() in it)

Could you please elaborate this in more detail?

Regards

Rüdiger


Re: svn commit: r678179 - /httpd/httpd/branches/2.2.x/STATUS

Posted by Nick Kew <ni...@webthing.com>.
On Mon, 21 Jul 2008 08:51:54 -0400
"Eric Covener" <co...@gmail.com> wrote:

> On Mon, Jul 21, 2008 at 8:37 AM, Nick Kew <ni...@webthing.com> wrote:
> > On Mon, 21 Jul 2008 08:09:23 -0400
> > "Eric Covener" <co...@gmail.com> wrote:
> >> Is it kosher to pull the API extension from trunk? A less intrusive
> >> (AuthnProviderAlias-only) solution occurred to me while reading
> >> your note.  Would the (incorrect) minor bump be left in place or
> >> can it be reverted?
> >
> > I'd say it wants a major bump that'll draw the line under likely
> > breakage.  Fortunately that's allowed in /trunk/.
> >
> 
> For clarification, you're saying a major bump to remove
> has_realm_hash() right in favor of a simpler solution?

Yep.  Since you have a solution that doesn't break the API,
use it and revert the API change.  The major bump draws a
line under any incompatible usage that may have happened
in the meantime.

> The following [chop]

Not reviewing that now, but will try and find time for it
if you implement it.

-- 
Nick Kew

Application Development with Apache - the Apache Modules Book
http://www.apachetutor.org/

Re: svn commit: r678179 - /httpd/httpd/branches/2.2.x/STATUS

Posted by Eric Covener <co...@gmail.com>.
On Mon, Jul 21, 2008 at 8:37 AM, Nick Kew <ni...@webthing.com> wrote:
> On Mon, 21 Jul 2008 08:09:23 -0400
> "Eric Covener" <co...@gmail.com> wrote:
>> Is it kosher to pull the API extension from trunk? A less intrusive
>> (AuthnProviderAlias-only) solution occurred to me while reading your
>> note.  Would the (incorrect) minor bump be left in place or can it be
>> reverted?
>
> I'd say it wants a major bump that'll draw the line under likely
> breakage.  Fortunately that's allowed in /trunk/.
>

For clarification, you're saying a major bump to remove
has_realm_hash() right in favor of a simpler solution?

> I should add: my first thought on reviewing this was that has_* should
> be a bitfield so it could be generalised to other methods.  But I
> still don't see why just testing whether the get_realm_hash member
> is NULL isn't sufficient for you.

There was an existing test at initialization, but with
AuthnProviderAlias get_realm_hash() is always non-NULL.

The following didn't occur to me before the patch under review.  We
can confine the changes to AuthAliasProvider and still catch it at
initialization time (no real patch yet)

static const authn_provider authn_alias_provider =
{
    &authn_alias_check_password,
    &authn_alias_get_realm_hash,
};

static const authn_provider authn_alias_provider_nodigest =
{
    &authn_alias_check_password,
    NULL,
};

static const char *authaliassection(cmd_parms *cmd, void *mconfig,
const char *arg) {
...
ap_register_auth_provider( ...
                                  provider->get_realm_hash ?
&authn_alias_provider : &authn_alias_provider_nodigest,
...

This would satisfy the existing test performed in mod_auth_digest.

-- 
Eric Covener
covener@gmail.com

Re: svn commit: r678179 - /httpd/httpd/branches/2.2.x/STATUS

Posted by Nick Kew <ni...@webthing.com>.
On Mon, 21 Jul 2008 08:09:23 -0400
"Eric Covener" <co...@gmail.com> wrote:

> On Sat, Jul 19, 2008 at 12:03 PM,  <ni...@apache.org> wrote:
> 
> >  * mod_auth_digest: Detect during startup when AuthDigestProvider
> >    is configured to use an incompatible provider via
> > AuthnProviderAlias. @@ -164,6 +149,10 @@
> >    Backport version for 2.2.x of patch:
> >         http://people.apache.org/~covener/2.2.x-digest_vs_authnalias-2.diff
> >    +1: covener, rpluem
> > +   -1: niq: looks like a major MMN bump to me, in that it'll
> > +            break binary compatibility with old compilations
> > +            of authn modules (has_realm_hash being undefined).
> > +            Why not just test get_realm_hash - as in the else
> > clause?
> 
> Is it kosher to pull the API extension from trunk? A less intrusive
> (AuthnProviderAlias-only) solution occurred to me while reading your
> note.  Would the (incorrect) minor bump be left in place or can it be
> reverted?

I'd say it wants a major bump that'll draw the line under likely
breakage.  Fortunately that's allowed in /trunk/.

I should add: my first thought on reviewing this was that has_* should
be a bitfield so it could be generalised to other methods.  But I
still don't see why just testing whether the get_realm_hash member
is NULL isn't sufficient for you.

> (AuthnProviderAlias can just register a different authn_provider
> struct and not put a reference to its own own get_realm_hash() in it)

As Rüdiger just said, Hmm?

-- 
Nick Kew

Application Development with Apache - the Apache Modules Book
http://www.apachetutor.org/