February 2001 - Zope-Dev

LONGing for normal inetegers...
by Jon Franz 14 Feb '01

14 Feb '01

I had this problem in the past and hacked the mysql DA to fix it, then dicovered to my dismay I was using an out-of-date mysqlDA and it had already been fixed... Which DA are you using? >Message: 1 >Date: Sat, 10 Feb 2001 16:30:19 -0800 (PST) >From: francine broekhove <francine(a)burningmail.com> >To: zope-dev(a)zope.org >Reply-To: francine(a)burningmail.com >Subject: [Zope-dev] LONGing for normal integers ... > >Hi everyone, > >i've updated my Zope (finaly) and my Python and my Linux (have been bussy). But, trying to fire up my 'old' applications in Zope, >mostly connected to MySQL databases, i get an annoying 'L' behind every integer. > >All my indexes are integers. I could rewrite all my dtml to include _.int(integer) or something, but that seems completely stupid. >Isn't there a more 'central' way to solve this. F.i. should i adapt some python types? Does anyone know what is wrong and how to >fix this?

3 2

Re: [Zope-dev] Re: ThreadSafeCounter
by Tres Seaver 14 Feb '01

14 Feb '01

> "Andy McKay" <andym(a)ActiveState.com> wrote: > > > | The ideal solution would be to use an object that lives in the ZODB, > > | I wonder if there is a way to keep the 'object history' empty? That > > | is, keeping the counter 'packed' while retaining 'object history' > > | information on all other objects. > > That would work, however I just dont think that fits neatly into the ZODB in > any way since all objects are appended. Small confusion here: "ZODB" is a higher-level abstraction than "FileStorage"; it is quite feasible to have one or more storages inside your Zope which are "packless" / "undo-less", and which then have the space efficiencies you are looking for (often at a non-trivial time cost). Tres. -- =============================================================== Tres Seaver tseaver(a)digicool.com Digital Creations "Zope Dealers" http://www.zope.org

3 4

ThreadSafeCounter
by Andy McKay 13 Feb '01

13 Feb '01

So the only difference between this and FSCounter is that you implemented the locking I didn't. To cut a long story short the reason i didnt implement locking was Python on Windows doesnt have fnctl (http://velocity.activestate.com/docs/ActivePython/lib/module-fcntl.html) Perl does though.... If anyone knows of a module to lock and unlock cross platform I would appreciate it. Zope by the way locks in /lib/python/ZODB/lock_file.py, but there is no unlock. Maybe if we added an unlock in here this would provide a global lock and unlock function. I released FSPoll recently and was going to combine the two into one FSCountThing with FSPoll and FSCounter subclassing of it, so maybe we could co-operate on ThreadSafeCounter and FSCounter? Cheers. -- Andy McKay.

7 10

Re: [Zope-dev] ProxyPass and SiteAccess getting REMOTE_ADDR
by Oliver Bleutgen 13 Feb '01

13 Feb '01

> From: Oliver Bleutgen <myzope(a)gmx.net> >> Btw. with a small module it's even possible to log the usernames with > apache, >> something medusa isn't capabable of afaik (as of zope version < 2.3). > Cool! What module is this, and how do you use it? Heh, it's a apache-module written in perl called AuthAny that I leeched somewhere and it's sole purpose is to make apache think there's authentication going on so that he logs the names and to _require_ a username and password, but I think that is easily fixable if you want to allow anonymous logins. I would describe that as a hack, but I wouldn't be able to modify apaches itself for a more "correct" solution even when my life depended on it. I use this in http.conf <Directory proxy:*> PerlAuthenHandler Apache::AuthAny AuthType Basic AuthName "Your Login..." AuthUserFile "/web/site/bogus_userfile.txt" require valid-user </Directory> Don't rember whether AuthUserFile is necessary, it may be just an artifact of some other experiments. AuthAny.pm is an amazingly complex (;-)) perl-module which has to be put into the right location, which is in Suse (7.0): /usr/lib/perl5/site_perl/5.005/i586-linux/Apache/AuthAny.pm package Apache::AuthAny; # file: Apache/AuthAny.pm use strict; use Apache::Constants qw(:common); sub handler { my $r = shift; my($res, $sent_pw) = $r->get_basic_auth_pw; return $res if $res != OK; my $user = $r->connection->user; unless($user and $sent_pw) { $r->note_basic_auth_failure; $r->log_reason("Both a username and password must be provided", $r->filename); return AUTH_REQUIRED; } return OK; } 1; __END__ One note: Does anybody feel inspired to write a python module for apache which hooks into zope's security mechanism and gives a possibility for simple authentication which is shared between zope and apache? For instance authenticate a user iff he is allowed to access a certain "dummy" object. That way one could easily manage access within zope to files/scripts which are not inside the ZODB. cheers, oliver

1 0

Formal argument for acquisition and aq_self
by Steve Alexander 13 Feb '01

13 Feb '01

I've just been reading through some academic papers by Tom Mens. The paper at this link looks interesting, as it appears to me to describe a problem that is addressed in Zope by acquisition wrappers. http://prog.vub.ac.be/Research/ResearchPublicationsDetail2.asp?paperID=51 Abstract Most object-oriented languages provide a self pseudo variable that serves two purposes: invoking methods through self sends and returning a self reference. Current formal models capture this behaviour by introducing a self parameter that fulfils both roles. We will show that this double functionality is incompatible with object-based encapsulation. Moreover we will point out that the separation of both functionalities is needed in type systems. We propose a formal model based on a primitive self send construct, and present its denotational semantics. A self reference is implemented through a self send instead of the other way around. This model is extended formally with an object-based encapsulation mechanism. It allows us to keep inheritance, encapsulation and typing orthogonal. This way we obtain a flexible and incremental software development scheme. In the case of Zope, the "self sends" are represented by the way an acquisition wrapper dispatches messages. The "self reference" is given by aq_base (or one of the other aq_ attributes, depending on what you want "self" to mean at the time). -- Steve Alexander Software Engineer Cat-Box limited http://www.cat-box.net

1 0

Windows registry?
by Andy McKay 13 Feb '01

13 Feb '01

Anyone tried or got a script to create / delete the windows registry settings for running a zope server as a service. It would make my life a lot easier... If no-one has I could recreate them from observation. Cheers. -- Andy McKay.

3 3

Strange AttributeError in AccessControl/User.py
by morten＠esol.no 13 Feb '01

13 Feb '01

Hi guys, indeed a strange problem, anyone experienced this? Zope Error Zope has encountered an error while publishing this resource. AttributeError Sorry, a Zope error occurred. Traceback (innermost last): File /home/morten/Zope/lib/python/ZPublisher/Publish.py, line 222, in publish_module File /home/morten/Zope/lib/python/ZPublisher/Publish.py, line 187, in publish File /home/morten/Zope/lib/python/Zope/__init__.py, line 221, in zpublisher_exception_hook (Object: CatalogAware) File /home/morten/Zope/lib/python/ZPublisher/Publish.py, line 162, in publish File /home/morten/Zope/lib/python/ZPublisher/BaseRequest.py, line 442, in traverse File /home/morten/Zope/lib/python/AccessControl/User.py, line 496, in validate (Object: RoleManager) File /home/morten/Zope/lib/python/AccessControl/User.py, line 231, in allowed AttributeError: aq_inContextOf Thanks. -Morten

3 2

Python Scripts
by Chris Withers 13 Feb '01

13 Feb '01

Evan Simpson wrote: > > You shouldn't have to do anything special; You just call it. How would I pass a DTML namespace that I might have lying around when I make the call so that it gets bound to the appropriate Namespace variable in the script? cheers, Chris

1 0

Re: [Zope-dev] ProxyPass and SiteAccess getting REMOTE_ADDR
by Oliver Bleutgen 13 Feb '01

13 Feb '01

> From: "Chris Withers" <chrisw(a)nipltd.com> >> We're actually phasing this hack out in favour of a Virtual Host Monster > which >> seems like a much cleaner solution... > Sorry, Chris, VHM is irrelevent to this problem. If you want to know the > original remote IP, you have two choices: > 1. Use one of the Apache patches that have been posted. > 2. Mangle it into URLs and extract it with an AccessRule. > Neither one of these is really satisfactory at the moment, since medusa > doesn't know either solution, and so z2.log doesn't get the proper IP > addresses. I'm hoping that one of these days we can standardize on #1 and > provide support in medusa. I don't get why it is necessary at all to use medusa's logging when using apache or squid etc. as a proxy. I for instance use apache to also serve content which isn't contained in the ZODB (large files etc.), and I want the logs/statistics to also contain information about these. Btw. with a small module it's even possible to log the usernames with apache, something medusa isn't capabable of afaik (as of zope version < 2.3). cheers, oliver

2 1

Virtual Host Monster Paranoia
by Chris Withers 13 Feb '01

13 Feb '01

Right, I really like the idea of these things but I am concerned about something that allows anonymous users to futz with traversal. Can someone put my fears to rest that using these won't let anonymous users do bad things to my sites? cheers, Chris

5 7