April 1999 - Zope-Dev

ZServer design ideas: gneeral comments
by Ross J. Reedstrom 12 Apr '99

12 Apr '99

Hey all! I recently returned from the IEEE Metadata '99 meeting, and had some thoughts that I nned to share with the Zope developers. First off, I did a little evangelism, recomending scripting in general and Zope/Python in particular where ever I could :-) One particular case involves a software/data archive at a NASA. Under the title "Smart Data, Dump Archive", they've come up wit some thing they call 'buckets' that serve to contain data and associated software to act on that data, that can be moved around between archives, and is accessed through the web. Hmm, sounds like WWW enabled objects with methods, doesn't it? Where have I heard this before? They are currently hand crafting perl/CGI to do all this, so I recommended a look-see at Zope. The one piece they're looking for that Zope doesn't do well right now is the packaging everything up and moving it around part. Mostly because of the 'fear of pickles' issue - err, I mean 'fear of arbitrary foreign code'. A second point from the meeting was about interoperability. The general concensus was "Standards, Standards, Standards, oh and then some more Standards". Admitedly, there's more standards than you can shake a stick at covering all aspects of something like this, but the one application I though of was ZServer. My understanding is that it is ZServer's job to provide different "views" of the ZopeBase (err, what was the agreed on term for this?). One of design issues around the FTP view is what to do with attributes and methods of docs. One suggestion was an XML doc, containing the extra info. All that 'extra info' is what the whole meeting was about, namely Metadata! There exists now a proposal to the OMG (Object Managent Group - the CORBA people) to fold aspects of UML into XML and call it XMI - XML Metadata Interchange. It's a little verbose (what commitee standard isn't?) but there are working implementations in use in the database world, allowing exchange of object schemas and such between different vendors tools. Take a look at: http://www.oasis-open.org/cover/xmi.html an archived version of the announcement of the modified proposal: http://www.omg.org/archives/orbos/msg00702.html To quote IBM (one of the submittees) on the proposal: "The main purpose of XMI is to enable easy interchange of metadata between modeling tools (based on the OMG UML) and between tools and metadata repositories (OMG MOF based) in distributed heterogeneous environments. XMI integrates three key industry standards: 1) XML - eXtensible Markup Language, a W3C standard; 2) UML - Unified Modeling Language, an OMG modeling standard; 3) MOF - Meta Object Facility and OMG modeling and metadata repository standard." >From Zope's point of view, it provides an standard DTD creating an XML representation of the ZopeBase. HTH, Ross -- Ross J. Reedstrom, Ph.D., <reedstrm(a)rice.edu> NSBRI Research Scientist/Programmer Computer and Information Technology Institute Rice University, 6100 S. Main St., Houston, TX 77005

2 3

What does a "no Content-Type" traceback indicate is wrong?
by skip＠mojam.com 12 Apr '99

12 Apr '99

I'm looking for a little advice. I stuck a little sendmail block at the start of Response._traceback: def _traceback(self,t,v,tb): if t != "NotFound": import traceback, os msg = os.popen('/usr/sbin/sendmail -t', 'w') msg.write('To: skip(a)musi-cal.com\n') msg.write('Subject: Z traceback\n\n') traceback.print_tb(tb, None, msg) msg.write("%s: %s\n" % (t, v)) msg.close() tb=self.format_exception(t,v,tb,200) tb=join(tb,'\n') tb=self.quoteHTML(tb) return "\n%s\n%s\n%s" % (_tbopen, tb, _tbclose) I expect to use this mostly as a sanity check on my own changes. (If I introduce any bugs, I get a bunch of email. How apropos...) There are three types of tracebacks popping up frequently in my email at the moment. One is the ubiquitous "NotFound" error (which, as you can see above, I toss). A second common traceback indicates a user failed to fill in one or more required fields in a form. I will probably eliminate it from considerations as well. The third traceback I see frequently looks like File "/home/killer-whale/skip/src/Zope-1.9.0-src/lib/python/ZPublisher/Publish.py", line 872, in publish_module publisher = ModulePublisher( File "/home/killer-whale/skip/src/Zope-1.9.0-src/lib/python/ZPublisher/Publish.py", line 163, in __init__ fs=FieldStorage(fp=fp,environ=environ,keep_blank_values=1) File "/usr/local/lib/python1.5/cgi.py", line 846, in __init__ ctype, pdict = parse_header(self.headers['content-type']) File "/usr/local/lib/python1.5/cgi.py", line 690, in parse_header plist = map(string.strip, string.splitfields(line, ';')) exceptions.TypeError: argument 1: expected read-only buffer, None found This indicates the input was missinf a Content-Type header. Are there any valid circumstances under which this might occur? Does this just represent a broken browser or communication-challenged connection? Can I fake a content type and continue or is it a lost cause at this point because something more tragic will likely happen further down the line? Thx, Skip Montanaro | Mojam: "Uniting the World of Music" http://www.mojam.com/ skip(a)mojam.com | Musi-Cal: http://www.musi-cal.com/ 518-372-5583

2 2

RE: [Zope-dev] Minor Fix to CopySupport.py
by Brian Lloyd 12 Apr '99

12 Apr '99

> The current CVS's CopySupport._get_id funtion is a little buggy. > When pasting a copied object more than one time, the generated > Ids are > > [original] > [copy of original] > copy2 of copy of original > copy3 of copy2 of copy of original > ... > > which is misleading. The attached function code (sorry, > I have no diff available) fixes this to > > [original] > [copy of original] > copy2 of original > copy3 of original > ... > > and contains a proposal also: The first two lines > > try: return ob._get_id(id) > except AttributeError: pass > > call a container's _get_id method if available to allow customization. Thanks - I've added this for 1.11.0 Brian Lloyd brian(a)digicool.com Software Engineer 540.371.6909 Digital Creations http://www.digicool.com

1 0

Product API tutorial files?
by kent＠tiamat.goathill.org 12 Apr '99

12 Apr '99

There are some example files described in the Product API tutorial which I can't locate at the Zope site. The tutorial doesn't explain a few things and I suspect the tutorial programs would answer some of my questions. Anyone seen the files? Thanks Kent

1 0

Re: [Zope-dev] RE: [Zope] RE: No Access file found.
by Chris Larson 12 Apr '99

12 Apr '99

On 4/12/99 9:58 AM , Scott Robertson sroberts(a)codeit.com said: >> > Anyone? Anyone? Bueller? >> > >> > OK. Allow me to clarify. Is there any way to redirect the >> > path for the >> > 'access' path when it should be somewhere other than >> > INSTANCE_HOME? Say, >> > for example, I have multiple instances of Zope databases, >> > with one Zope >> > install, and I want to use a single superuser, or access >> > file. Not what >> > Im' doing, but closer to a mebbe-real-world use. As far as I >> > can tell, >> > Zope wants 'access' located at '../INSTANCE_HOME'. Any >> > ideas? Or am I >> > just wrong (And/Or, that is). >> 'fraid to tell ya that INSTANCE_HOME is hard wired into the code. We're >> not likely to change that lest we want to get the 'evil eye of Jim'. >> But seriously, it is a pretty good idea to have something like this in >> one place, for us at least, not you. >> >> I cannot therefore recommend that you modify your Zope. I also cannot >> recommend that you would modify it in lib/python/AccessControl/User.py >> in the Super class around line 270 or so. >I've been meaning to bug you guys (DC) about this. Some how, and I'm not >really sure how it happened, maybe it was hard drive glitch or something like >that caused a line in _my_ lib/python/AccessControl/User.py to change from > f=open('%s/access' % INSTANCE_HOME, 'r') to > f=open('%s/access' % SOFTWARE_HOME, 'r') >and ever since that happened I've been able to manage multiple instances >of zope, each with their own database, with one superuser account. Oh and >I belived the access file moved to zope/lib/python as well, must be >related to that same hard drive glitch I was talking about. It was freaky, >but turned out to be really useful if you know what I mean. Scott, Sorry to hear about your troubles. Off hand, it sounds like you got hit with the "SOFTWARE_HOME" variant of Melissa. I am right now going to make absolutely sure no such horrific occurrence occurs in my own Zope installation. You done a great service in warning the Zope community about this malicious trojan. I had inadvertently caused the very thing Michel warned me about in his email, but your unfortunate predicament sounds like something that I should be even more careful about avoiding. Thanks, (and grins), Chris Larson PS For anyone out there who has witnessed my travails with my ISP, NFS, etc, once I got var/ up an running under /tmp and set up a shell script to periodically backup and refresh the /tmp/var directory, and lamentably noticed the changes in User.py, everything is comin' up roses! I am now Zopin' away. And on the off chance noone has caught on to this little tidbit, Zope _rocks_!

1 0

Minor Fix to CopySupport.py
by Stefan Franke 12 Apr '99

12 Apr '99

The current CVS's CopySupport._get_id funtion is a little buggy. When pasting a copied object more than one time, the generated Ids are [original] [copy of original] copy2 of copy of original copy3 of copy2 of copy of original ... which is misleading. The attached function code (sorry, I have no diff available) fixes this to [original] [copy of original] copy2 of original copy3 of original ... and contains a proposal also: The first two lines try: return ob._get_id(id) except AttributeError: pass call a container's _get_id method if available to allow customization. * * * If you change CopySupport.py please have a look to "[Zope-dev] Possible bug in CopySupport.py (Zope 1.10.2)" as well, which I posted two weeks ago to zope-dev Thanks, Stefan * * * [CopySupport.py, line 426] def _get_id(ob, id): try: return ob._get_id(id) except AttributeError: pass try: ob=ob.aq_base except: pass n=0 if (len(id) > 8) and (id[8:]=='copy_of_'): n=1 orig_id = id while (hasattr(ob, id)): id='copy%s_of_%s' % (n and n+1 or '', orig_id) n=n+1 return id

1 0

RE: [Zope] RE: No Access file found.
by Michel Pelletier 12 Apr '99

12 Apr '99

> -----Original Message----- > From: Chris Larson [mailto:clarson@changeling.com] > Sent: Tuesday, April 06, 1999 2:32 PM > To: Zope Mailing List; Zope Dev List > Subject: [Zope] RE: No Access file found. > > > Anyone? Anyone? Bueller? > > OK. Allow me to clarify. Is there any way to redirect the > path for the > 'access' path when it should be somewhere other than > INSTANCE_HOME? Say, > for example, I have multiple instances of Zope databases, > with one Zope > install, and I want to use a single superuser, or access > file. Not what > Im' doing, but closer to a mebbe-real-world use. As far as I > can tell, > Zope wants 'access' located at '../INSTANCE_HOME'. Any > ideas? Or am I > just wrong (And/Or, that is). > 'fraid to tell ya that INSTANCE_HOME is hard wired into the code. We're not likely to change that lest we want to get the 'evil eye of Jim'. But seriously, it is a pretty good idea to have something like this in one place, for us at least, not you. I cannot therefore recommend that you modify your Zope. I also cannot recommend that you would modify it in lib/python/AccessControl/User.py in the Super class around line 270 or so. -Michel > Or any ideas for getting around NFS problems without putting > '/var' in > /tmp? > > Thanks, > > Chris > > >OK. After finally getting my ISP to migrate me to a server > with a current > >Python, I once again leap into the fray. I've placed my 'var/' into > >'/tmp/', to get around NFS, and updated my instance home to > /tmp. Now I > >have a shiny new problem to play with. Zope can't see > 'access'. Is the > >path to access assumed from INSTANCE HOME, or is there some > place I can > >specify it? (I'd kinda rather avoid putting it in /tmp, ya > know?). And > >before ya ask, yes there _is_ an access file in > >/---my-path---/cgi-bin/Zope-1.10.2-src/, and it does contain a > >"superuser:" style entry. I've included the traceback, and > my Zope.cgi below. > > > >Any help would be much appreciated. > > > >Chris > > > >PS Any comments on using '/tmp' to contain 'var/' as an NFS > bypass? '/tmp' > >is not mounted via NFS. > > > > The traceback > >----------- > >Sorry, a SERIOUS APPLICATION ERROR occurred.<p> > > > > > > > >My 'Zope.cgi' > >----------- > >#!/---my-path---/cgi-bin/Zope-1.10.2-src/pcgi/pcgi-wrapper > >PCGI_NAME=Main > >PCGI_MODULE_PATH=/---my-path---/cgi-bin/Zope-1.10.2-src/lib/p > ython/Main.py > >PCGI_PUBLISHER=/---my-path---/cgi-bin/Zope-1.10.2-src/pcgi/pc > gi_publisher.py > >PCGI_EXE=/usr/bin/python > >PCGI_SOCKET_FILE=/tmp/var/pcgi.soc > >PCGI_PID_FILE=/tmp/var/pcgi.pid > >PCGI_ERROR_LOG=/tmp/var/pcgi.log > >PCGI_DISPLAY_ERRORS=1 > >BOBO_REALM=Zope > >BOBO_DEBUG_MODE=1 > >INSTANCE_HOME=/tmp > > _______________________________________________ > Zope maillist - Zope(a)zope.org > http://www.zope.org/mailman/listinfo/zope > > (For developer-specific issues, use the companion list, > zope-dev(a)zope.org - http://www.zope.org/mailman/listinfo/zope-dev ) >

2 1

[Zope] The big 3, and Zope initiatives for Linux Expo
by jhauser＠ifm.uni-kiel.de 12 Apr '99

12 Apr '99

Great news about things on the horizon. Probably many of the current zope-developers have in part the same goals, not a complete Portal in the beginning though. Here are some questions regarding some subgoals. I'm mainly interested in the ``Personal Areas''. Brian Lloyd writes: > We plan to demonstrate this by reinventing the zope.org site, > using it as a demonstration and case-study of a portal site > that incorporates: > > o Powerful information browsing and searching > > o News items > > o Issue tracking Means this calendar and schedule? I thought of this as a good area where acquisition could be used. The appointments for a department can be acquired by the calendars of the employees. And the employees can make their personal schedule with a folder-method of the Personal area folder. > o Community membership Is this only achieved with permissions, or are there special methods involved? > > o Community projects > > o Personal areas that community members can maintain themselves I think this will be a folderish object, right? I try to implement something like this. But it is hard to make this general, because there is currently no easy method, besides external methods to extent the functionality of such products. Will it be possible to extent products in zope, without making a new one. The problem now is, that external methods need to be deployed at possible many locations. > > o Community involvement in the form of polls and discussions > but anyway, great to see all these new possibilities. Thanks Zopistas :-) __Janko

1 0

Re:[Zope] help with magic method "__of__"
by Butch Landingin 12 Apr '99

12 Apr '99

Hi, I've been playing around with Confera for about 4 weeks now -- but I'm not the original developer nor am I an employee of Digital Creations (hint!hint!) so <YMMV>take the following with a grain of salt</YMMV>. On April 11, 1999 "Karl & Mel" <llwo(a)dbtech.net> wrote: >I know that Confera is not supported, but I am using it as a base for an = >extended version that would allow messages to be moved, reordered, = >copied, paseted, etc. In otherwords turn Messages into folderish? = >containerish? objects. I don't want to give up the indexing and = >searching though. Which leads me to believe that I need to override the = >'setItem' and '__getitem__' maybe? I think you'd probably need to add methods in the Message/Topic classes to remove a child message (i.e. id in the intSet self.ids) (this would be to implement a <cut> function) and reattaching itself to another message or topic (to implement a paste function) You can *probably*(just a guess, I haven't fiddled around too much with this) use the setitem to paste messages to another message but it does'nt have to do the indexing because it would already have been indexed when it was added, but it also has to handle reorganizing the thread field of the message you are moving so that these reflect the new paths (note that if the message being moved has children (i.e. replies), their thread fields also have to be updated and so on recursively) I don't know why you'd have to modify __getitem__ though. In the cut function, you can probably base it on the delItem function (just take out the unindexing and delete object parts) As for the other functions (reodering,etc.), I don't have enough info to suggest anything... > >Anyway > >As I've been reading the code over and over and over (two weeks now) and = >have thrown out three attempts. I keep seeing things like > >self.data[id].__of__(self) >msg=3Dmsg.__of__(self) >r=3Dmap(lambda x, p=3Dself: x.__of__(p), r) > >How does the magic method '__of__()' work? > >Also how does intSet{} work? > >ta for any info. Grab the code in the Zope CVS area. If you're on Windows, theres a nice gui front end called WinCvs (http://www.wincvs.org) that allows you to mirror the entire Zope source distribution on your local drive. The reason for this is that the definitions of IntSet and the magic __of__ methods are defined in the c modules which are included in the source distribution (see Acquisition.c and intSet.c). AFAIK, obj.__of__(self) attaches your object (obj) to the Acquisition hierarchy (of which the self object is already a part of) so that it becomes accessible as a URL in Zope ...In Zopish, this is known as '*wrapping* the object into an acquisition wrapper'... If this is not as clear as it should be it's because I'm not that exactly sure of it myself :( Maybe somebody from DC or the original Confera developer can confirm this... HTH. Butch Landingin P.S. If you do get around to implement those things -- maybe you can send me some of the source code to see. I'm trying to finish up 0.1 version of a Zope Product that's based on Confera that implements a SlashDot-like website and I might be able to use some of the code you have <ain't Open Source great? ;)>. _________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com

2 1

RE: [Zope-dev] Bug in DateTime arithmetic
by Brian Lloyd 12 Apr '99

12 Apr '99

> Hello, everyone. > > While messing around with Zope, I came across a bug in the > DateTime class. > The following python session exhibits the faulty behavior: > > >>> date = DateTime('05/01/1999').earliestTime() > >>> d = (date - date.day()).earliestTime() > >>> print d > 1999/04/30 00:00:01 US/Eastern > >>> print (d - 24) > 1999/04/06 00:00:01 US/Eastern # correct > >>> print (d - 25) > 1999/04/05 00:00:01 US/Eastern # correct > >>> print (d - 26) > 1999/04/03 23:00:1 US/Eastern # incorrect > >>> d = d - (d.day() - 1) > >>> print d > 1999/03/31 23:00:1 US/Eastern # incorrect > >>> ^D > > Apparently, the DateTime code is picking up an extra hour > somewhere which > is throwing the calculation off. At first glance, I'd guess that you are seeing the (backward) transition from daylight savings time to standard time... Brian Lloyd brian(a)digicool.com Software Engineer 540.371.6909 Digital Creations http://www.digicool.com

2 1