- Zope-Dev - Zope lists

Re: [Zope-dev] CST not really... :-)
by Anthony Baxter 30 Aug '01

30 Aug '01

>>> Chris McDonough wrote > Though it would be kind of neat to play around with persistent > connections in conjunction with session tracking, it's not too high on > the list right now for CST.. that's not to say, however, that an > alternate session id manager could not be written to do this. > > I thought, however, that persistent HTTP connections were meant more to > avoid TCP connection costs than to identify sessions. That's right. Using a persistent connection to identify a session is way way bogus. In particular, note that _many_ proxies do not support persistent connections properly, if at all. Anthony

1 0

Re: [Zope-dev] CST not really... :-)
by Oliver Bleutgen 30 Aug '01

30 Aug '01

> Hi everyone, > But we think a really good CORE session tracking should be transparent > and independent of this cookies/forms options. This means be able > to install zope and have somewhere an option to turn on/off > zope "really core session tracking". This means zope > having the ability to do sessions using http1.1 persistent connections > which > medusa allready implements and also most current > browsers. Then no need to generate Tokens and pass them with cookies > or forms, there is allready a unique identifier between the server > and the client and that is in medusa socket_map. Nope, persistent connections are _not_ unique. RFC 2616 proposes that the client SHOULD implement no more then 2 persistent connections (i.e. it can use more than 1). And nowhere do I find a guarantee that the requests from the client have to use always the same persistent connection - i.e. they may close that connection and reopen another one. And how does the client know when to close the connection - it surely will not hold the connection open as long as the browser runs. So how does one handle the user which goes to a side, takes a break of 10 minutes and the continues there? Another data-point, there is an rfc, 2965, which explicitly deals with HTTP state management, and it doesn't mention http/1.1 persistent connections I think. cheers, oliver

2 1

xml-rpc/soap errors
by Andy McKay 29 Aug '01

29 Aug '01

Anyone got an elegant way of catching xml-rpc/soap errors so that they dont go to standard_error_message and return a whole bunch of html to the client as the error value? I really can only think of a <dtml-if> in standard_error_message, which doesn't count as elegant. Cheers. -- Andy McKay.

6 14

RE: [Zope-dev] CST not really... :-)
by sean.upton＠uniontrib.com 29 Aug '01

29 Aug '01

I really like the idea of persistence in the HTTP transport, but Chris is right... Keepalives are not at all reliable through proxies, and any well designed app that can take advantage of it ideally should be designed to be cache-enabled and cache-safe, not to mention proxy safe. The very fact that most corporate users go through trans-proxies shoots the persistent connection idea down, not to mention that it means that one cannot use a caching http accelerator like Squid or mod_proxy. The burden of session tracking should be at the application level, or just below, not at the transport level. HTTP is a stateless protocol, which is something we are going to have to live with, unless persistent connection id HTTP headers are supported by all browser clients and at least Squid, if not all the commercial proxies... Something along those lines would have to be an official ratified w3c recommendation with major vendor support for that to happen any time soon. Such a standard would have to support maintaining an _application_ session across multiple _transport_ sessions. If something like this was made possible in the future, it might be very good for XML-over-HTTP RPC protocols like SOAP and XML-RPC, but for now... The best we have for the practical time being is going to have to be cookies, etc... Sean -----Original Message----- From: Chris Withers [mailto:chrisw@nipltd.com] Sent: Wednesday, August 29, 2001 1:44 PM To: JXlio Dinis Silva; zope-dev(a)zope.org Subject: Re: [Zope-dev] CST not really... :-) > Any comments? someone thought about this before? someone from CST > development thought or is thinking on this possibility, i.e, zope > really core session tracking using http1.1 and no need for > cookies/forms/url_rewriting? What happens when your persistent connection breaks for some reason beyond your control? What about users going through caching proxies? cheers, Chris _______________________________________________ Zope-Dev maillist - Zope-Dev(a)zope.org http://lists.zope.org/mailman/listinfo/zope-dev ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope )

1 0

CST not really... :-)
by J�lio Dinis Silva 29 Aug '01

29 Aug '01

Hi everyone, We have been looking at Core Session Tracking. >From the name of this product we thought this means this product enabled zope to track zope requests grouping some requests has a unique session, i.e, some way of zope tracking all requests from a specific client/browser/session. But no, this is a session managment product, i.e, data_manager/container and a programming api. The core doesnt stands for a code leaving on zope core, i.e, somewhere between Zserver and Zpublisher, but this word core is more related to the physical place where sessions data are stored which goes from standard FileStorage to BerkeleyStorage on a ZODB3 model basis. There were/are other alternatives SQL, FileSystem, RAM, etc. This product offers a good api for one to implement sessions using either cookies, url rewriting or form hidden fields, and it is allready prepared to scale in a zeo cluster. But we think a really good CORE session tracking should be transparent and independent of this cookies/forms options. This means be able to install zope and have somewhere an option to turn on/off zope "really core session tracking". This means zope having the ability to do sessions using http1.1 persistent connections which medusa allready implements and also most current browsers. Then no need to generate Tokens and pass them with cookies or forms, there is allready a unique identifier between the server and the client and that is in medusa socket_map. What we think is that with zope one dont need to use cookies, or url rewriting, or forms hidden fields. We should be able to simply call a method in a standard_html_header which activates sessioning. And no need for cookies, or forms, or heavy code writing. We have actually made a small test using zserver/medusa/asyncore socket_map which is a list of all persistent connections (open zchannels) in a zope server. Of course there are technical considerations to implement a consistent solution but we have made a consistent proof-of-concept. Any comments? someone thought about this before? someone from CST development thought or is thinking on this possibility, i.e, zope really core session tracking using http1.1 and no need for cookies/forms/url_rewriting? Best Regards, JS PS.For reference: See the section "A Better Solution to Session Tracking: HTTP/1.1" on: http://www.internettg.org/newsletter/mar00/workshop_session_management.html _________________________________________________________________ Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp

2 1

RE: [Zope-dev] Extending a class' __bases__ at runtime
by Adrian Hungate 29 Aug '01

29 Aug '01

I have just checked that 4.1.2 tarball and it does contain directories, is this a compatibility thing? I used the python/zope tar.gz code to create the file under windows (Part of the same PatchKit). My linux box is currently very ill so I can only test under windows at the moment. Adrian... -- Adrian Hungate Try Zope - http://www.zope.org - You might like it. If you do, try my Zope products - http://www.zope.org/Members/haqa - You might like them too! -----Original Message----- From: Morten W. Petersen [mailto:morten@thingamy.net] Sent: 29 August 2001 18:12 To: Adrian Hungate Cc: R. David Murray; Chris Withers; zope-dev(a)zope.org Subject: RE: [Zope-dev] Extending a class' __bases__ at runtime On Tue, 28 Aug 2001, Adrian Hungate wrote: > I think I must have missed the begining of this one, what is it you are > failing to achive? I have a hotpatch product that does what I think you > are talking about, so I guess we are not talking about the same thing. The PatchKit product? Doesn't seem like that'll do it no... (that's kinda cool by the way). Running 'tar zxfv <filename>' on the latest .tar.gz extracts them all as single files (without a directory) on Linux.. Cheers, Morten

1 0

get permission
by Johan Beauce 29 Aug '01

29 Aug '01

Hello, I use "manage_permission" for init permission of my directory. Afterthat, I need to show the permissions for this directory! Example : myFolder.manage_permission('View',listOfRoles,1) Now, I need to get permissons from myFolder ! What is the function which give me the permissions from myFolder ; a function like myFolder.getPermissions('View',ListOfRoles) ; that give me the list of Roles cheked from the permission View. Can you help me ? Regards, Johan.

2 1

RE: [Zope-dev] Extending a class' __bases__ at runtime
by Adrian Hungate 29 Aug '01

29 Aug '01

I think I must have missed the begining of this one, what is it you are failing to achive? I have a hotpatch product that does what I think you are talking about, so I guess we are not talking about the same thing. Adrian... -----Original Message----- From: Morten W. Petersen [mailto:morten@thingamy.net] Sent: 28 August 2001 22:53 To: R. David Murray Cc: Chris Withers; zope-dev(a)zope.org Subject: Re: [Zope-dev] Extending a class' __bases__ at runtime On Tue, 28 Aug 2001, R. David Murray wrote: > There's another reason to want to know the answer to this: the ability > to hotpatch the base classes of an existing zope class instead of > having to modify the source code for that class. This came up > for the person who was trying to create a version of the CMF that > used ZPatterns...and I don't recall hearing a solution. > > Plus, it sure is a "surprise" that something that works in normal > python code doesn't work in Zope! I presume this is an ExtensionClass > issue, but I don't know enough to suss it out. I'd offer $50 of merchandise at <put your favourite online store here> to anyone who fixes this issue. Cheers, Morten _______________________________________________ Zope-Dev maillist - Zope-Dev(a)zope.org http://lists.zope.org/mailman/listinfo/zope-dev ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope )

2 1

Oracle Connection and ZSQLMethods
by alan runyan 29 Aug '01

29 Aug '01

I have some strange behavior that I can replicate happening on all my boxes (dev and staging). Its a huge problem for me. I am running Zope2.4 w/ DCOracle2 on Windows NT4.0. on development I installed Oracle Connection in the root /db_connection /folder /folder/sql_stmts /folder/sql_stmsts/all of my zsqlmethods are here when I copy Data.fs over to staging, I delete the db_connection and recreate it with appropriate db_conn string. but **all** my ZSQLMethods do not register the newly created db_connection(?) as a available Data source and I get this when I edit them: There are no SQL database connections. You need to add a Zope SQL database connection before you can edit a Zope SQL Method. but as stated, I've already added my Connection at the root. I can even copy/paste it into the folder with the ZSQLMethods, as well as add it from scratch in the ZSQLMethods folder and still get the same error. should I **not** be moving this in a Data.fs? I am also using Oracle StoredProcs and they are having the same problem. the crazy thing is -- I can 'Test' them and call them and they work. I also cannot ADD any new ZSQLMethods (no connections available) on the staging box. thanks, ~runyaga

3 2

Extending a class' __bases__ at runtime
by Morten W. Petersen 28 Aug '01

28 Aug '01

Hia, trying to modify a class at runtime in Zope is proving hard to do. The first interactive session is a 'Zope debugging session': Python 1.5.2 (#0, Apr 10 2001, 10:03:44) [GCC 2.95.3 20010219 (prerelease)] Copyright 1991-1995 Stichting Mathematisch Centrum, Amsterdam Welcome to <secure_monitor_channel connected 127.0.0.1:46914 at 90a7df8> >>> import Zope >>> a=Zope.app() >>> s=a['Secret'] >>> class a1:pass >>> class a2:pass >>> class a3:pass >>> s.__class__.__bases__ = s.__class__.__bases__ + (a1,a2,a3) >>> s.__class__.__bases__ (<extension class App.PersistentExtra.Persistent at 4022be80>, <extension class Products.BTreeFolder.BTreeFolder.BTreeFolder at 84fecf8>, <extension class Acquisition.Acquirer at 4024ac80>, <extension class AccessControl.Role.RoleManager at 82819f0>, <class Products.WarpFramework.product_base_item.product_base_item at 845e058>, <extension class Products.WarpFramework.base_item.base_item at 848f1b0>, <class Products.Secret.imports.imports at 8532fa0>, <class Products.Secret.initializations.initializations at 858ee28>, <class Products.Secret.trash_test.trash_test at 857e2d0>) >>> where, as you can see, trying to change the base classes doesn't work, while in an ordinary, 'plain python' interactive session, it works: Python 1.5.2 (#0, Apr 10 2001, 10:03:44) [GCC 2.95.3 20010219 (prerelease)] on linux2 Copyright 1991-1995 Stichting Mathematisch Centrum, Amsterdam >>> class a: ... pass ... >>> b=a() >>> class a1: pass ... >>> class a2: pass ... >>> class a3: pass ... >>> b.__class__.__bases__ () >>> b.__class__.__bases__ = b.__class__.__bases__ + (a1,a2,a3) >>> b.__class__.__bases__ (<class __main__.a1 at 8070db0>, <class __main__.a2 at 806ff80>, <class __main__.a3 at 80709f8>) >>> a.__bases__ (<class __main__.a1 at 8070db0>, <class __main__.a2 at 806ff80>, <class __main__.a3 at 80709f8>) >>> Any clues on what needs to be done to be able to modify a class' base classes at runtime? Thanks & cheers, Morten

3 5