Transcript from the Aug 5, 1992 meeting of Smalltalk users on irc. Due to a capture buffer error, the first third of the meeting is missing. Your not missing much. This file does include all (?) of the most interesting talk, IMO. - mjb - mjb@netcom.com MDI means Multiple Document Interface. This is becoming a kind of +standard interface framework for windows applications. The Windows +Filemanager and the Sysedit utility are MDI applications Any expieriences with using a Smalltalk system with a *Group* of +programmers ? Rolf: only painful ones! ST/V uses bubblesort for adding elements to SortedCollections! tmu: What!! We too. We had to write our own progrect environment to load and store +projects, and to administrate users Yes! A rewrite using QS made the whole system behave a lot better Rolf: you use large teams? There are ionly 3 of us - so we use ad hoc +techniques Evereybody loves his/her image the most. How do you bring all the +efforts together? No large teams, but Smalltalk is difficult to use for even us 4 users +if they are to work together basically our boss imposes his image on us. I have to re-tailor it each +time I collect it off him. the first thing to do is to collect ones work in small packages Which can be filed out and filed in without damaging the system With all objects created? It's a good thing to start with classes and methods only *** Portnoy has joined channel #programST hi portnoy Hi all Hi Portnoy hi portnoy Hi i have only little time, but enough to announce that Version 4.1 +reached me at last Portnoy - thanks for your description of NESTOR baz: i'm now back from my holiday I'm capturing this for posting to the archives, please don't discuss +proprietary or other info I will have to edit out later, thanx Will speaking too freely get us in trouble? not from me, I just don't want to spend the time editing stuff out Portnoy: I promised a description of our OODDBMS and hypertext system to +you - I have not forgotten 4.1 makes a good impression on me, seems like all my little wishes +were heared 4.1 Parkplace Smalltalk ? Martin: can you please tell me the internet numerical address of the +site where the discussiion is archived I *must* do some smalltalking now. bye *** Signoff: Baz (Leaving) rolf: yes I've heard rumours about Parkplace Smalltalk for OS/2 2.0. Any infos? *** Signoff: Deepak (Bad link?) sorry, i have to leave too *** Signoff: Portnoy (Portnoy) I just received my first issue of "The Smalltalk Report", kinda skimpy, I +was hoping for more What kind of topics do they cover? Rolf: One column is "the best from comp.lang.smalltalk" The lead article is "Implementing Peer Code Reviews in Smalltalk" - +discusses using Envy/Developer environment... next, "Quality assurance issues...",... "The Best of comp.lang.smalltalk", "Smalltalk idioms: Abstract",... "Getting Real: Creating subclasses"... "Lab Report: ST research at Univ of Fla."... product announcements & "Highlights", fini martinb: thanks, doesn't seem too interesting to me 28 pages including coves and ads what price this was the july/aug issue, hopefully suffering the summer blahs *** tmu has changed the topic on channel #programST to Discussing Smalltalk +related topics tmu can you mail me your QC for SortedCollection? Might work for +Smalltalk/V PM as will Rolf: yes <$70/yr 6 or 8(?) issues editors are John Pugh and Paul White The same people from the JOOPS ? including ? "Carleton University & The Object People" a SIGS Publications product The Envy/Developer environment - should i know that ? I'm always looking for development systems - if only to improve our own Rold I have just created an app which also includes a better version of +"includes:" using a binary method probably, seems to be the leading (only ?) tool for ST that provides +facilities for the development and maintenance of large software systems in +ST tmu: works for all Collections ? provides... martinb: Can you mail me any stuff about that, or give an address to +get some paper about it? Rolf: only SortedCollection have enough structure that can be exploited versioning, configuration, and release management capabilities, mechanisms +for coordinating and integrating development activities of multiple +developers rolf, what's your email addr breu@rog1.rog.rwth-aachen.de rolf: you should get mail in a minute... rolf: tell me if it worked for ST/PM Thanks. I'm new in networking. Hope my configuration will receive it! Rolf: if you can receive mail... also, along the same lines is "AM/ST" from Coopers&Lybrand/ ImageSoft, who +can be reached at 1-800-245-8840 or 516-767-9067, or mcdhup!image!info tmu: The basic collections of all Smalltalk/V dialects are rather +similar. I think it should work Rolf: all similar "hacked"?-) of course! martinb: what is mcdhup!...? Email address? yep a uucp email addr Anybody else interested in binary add and include for SortedCollections +in ST/V xxx? I have the code ready to send have you uploaded it to the archives? Rolf here is the ENVY?Developer info that I have... martinb: Thanks. Do you know any archive that might contain fixes of +Smalltalk/V PM on OS/2 2.0? contact Object Technology International, 1785 Woodward Drive, Ottawa, +Ontario, Canada, K2C 0P9 or at 613-228-3535 Today, I have removed a bug from our system, which could well be another +candidate for "the most frequent ST slips". Want to hear details? the only archives I know of are manchester and uiuc Rolf: conatct the Digitalk forum on Compu$erve, they have the latest +patches ready for downloading (if you can afford our telecom service.....) I don't have a compuserve access, sorry. Is Digitalk present anywhere +on news? Digitalk currently supports their products through their "Direct Connect" +and their forum on CompuServe DT doesn't have an IP at this time That's both not for free, I think, not even if I want to post a bug +fix! Rolf: no, but you can have Dan Goldman's mail address. Reachable thru +Internet. But he won't forward you the files I think I'm not sure if they still offer support through BIX tmu: What is that address? martinb: what is BIX, and how to access it? martin: Can you please give me the internet adress of the irc-archive +site (numerical!)? Rolf what I can do is do a catalog of all the DT files and post it to the +archive, how you get them is another matter, cause it's gonna cost someone to +get them checking... It would be even great to know, what's there, thanks I don't have the numerics, but the manchester site is: cs.man.ac.uk and the +u.s. site is st.cs.uiuc.edu (128.174.241.10 ?) BIX is the Byte magazine Byte Information Exchange ( McGraw Hill publications) martinb: BIX: How to access? *** tonyKwong has joined channel #programST internet or telekom? *** Mode change "+o tonyKwong " on channel #programST by ucsu.colorado.EDU martin: what does (128.174.241.10 ?) mean? Is it complete? I did not +find the irc script at manchester, BTW. lets see... it uses tymnet to access if I recall correctly, basically, BIX +is for members only, though anyone can join, pick up a Byte magazine to get +details OK, I think by bookshop has the Byte magazine tmu: that a numerical site address, i.e. ftp 128.174.241.10 which is the +"real" way to access ... using ftp st.cs.uiuc.edu gets translated to the numerical Where are all the Smalltalkers? Wasn't 10.00 PDT agood time for +everybody? Martin: I thought there was a digit mis I have a feeling that the morning hours are not too good for U.S. +participants hello, just came back from lunch, it's ~1:30pm on the east coast +of the US tmu, it's very possible the munber is wromg Tony, do you have the st archive site addresses (numerically)? I can give it a try. will look st.cs.uiuc.edu is 128.174.241.10 the first st-irc meeting was held in the evening, u.s. time, and +participation was better thanx tony maybe there should be 2 conferences a month one for Europeans and one for +Americans (north south whatever ;) ) If we could get better participation we might be able to induce industry +folks to attend to answers questions about there product etc. er, their product Great idea Maybe one should make a mailing-list where everybody can subscribe who +wants to know about the next event, topics, etc right now we haven't reached "critical mass" I post all relevant info to comp.lang.smalltalk I have heard about irc sessions with all the glamorous icons of +Smalltalk: Adele Goldberg, and, and Perhaps, Europe has'nt a large Smalltalk community I post trnacripts of thes confs to the u.s. st site, which should get +mirrored to manchester It is sometimes hard to follow all threads in comp.lang.smalltalk yes Adele and a v.p. showed up at the first conference Can you give the exact path and name of the file. It sometiomes takes a +few minutes to change directories the site is now moderated and all uploads initially go into a tmp directory +and then is moved to the "irc" directory by the moderators, probably +something like /pub/irc/ tmu: is Smalltalk/V Windows now a robust product? I tested 1.0 and 1.1 +and they didn't work for more than 10 minutes at my system Rolf: 2.0 makes a good impression, but with all those new features, +there surely will be some surprises left. Most of the UAE were created by +Windows 3.0 which behaves much better now. tmu: you're running Windows 3.0 or 3.1? I have 2.0 in one window at the moment switching constantly between this +and the irc window. If it crashes I wont have time to say goodby!-) 2.0 in a window ?? I would like to tell you about the bug I recently removed as a warning, +because it ould be a common failure. please do ST/V Windows 2.0 (not ParcPlace) I use to have some collections of objects which I handle as Dictionaries +with the objects name as the key and the object itself as the value. This +sounds a bit redundant, but redundancy is sometimes necessary fro performance +reasons. Well, I am leaving th OO paradigm with this method So: I could not rely on object identity when renaming objects. The +objects WERE renamed but not the KEY! This produced links where the nodes which were linked could not be +referenced anymore, making the structured object unable to display itself +properly in the editor. tmu: Can't you use symbols as names? So I had to enrich the renaming procedure by also removing the object +under its old key and readding it under its new name, which is more like +tuple oriented database hacking - not object oriented. I will rethink my +usage of Dictionaries! *** karl_m has joined channel #programST hi all tmu: Or make a subclass of String, which reacts on #== by returning the +result of #=? Symbols as names? I recently posted a long article concerning the usage +of symbols and how system performance was degrading significantly. Just take +a look at how many message sends are involved in referncing symbols, not to +say adding symbols. hi karl, do your program in Smalltalk? Not a problem with 200 symbols but with 2000 it gets hairy. ahh - now IC - i thought ST stood for atari ST - ohh well never mind +:) i program all right but in 'C',6502,8086,i860,88k,68k etc *** martinb has changed the topic on channel #programST to programming in the +Smalltalk language *** Bad Craziness (from toaster.sfsu.edu) *** Bad Craziness (from dreamtime.unm.edu) Rolf: you dont mean to do something like "become:"? thanx martin :) anyway see ya all l8r *** karl_m has left channel #programST bye :) I had already set the topic to something unambigious.. But this is better! apparently not ;) tmu: No. What's needed seems to be a 'String' class where the objects +are #== if they contain the same characters happens at every conf Rolf: that will not help, because there are TWO representations of ONE +object, the Dictionary's KEY (only the objects name) and the object itself, +the VALUE. So you always have to maintain at two locations. A solution would +be a new kind of == is equivalence (same object) so wouldn't == check pointers? (sorry novice +question) dictionary with a synchronized KEY-VALUE association and an appropriate +rename method You dont want to force ALL objects in the global image to change a +string XX they point at toi reference a new string, say YY tmu: Sorry, I misunderstood your problem. Rolf: so you have luckily not come across this problem. martinb: == is same object, but you may overwrite it We are using up to 10.000 objects with 6.000 links (which are themselves +quite complex "first-class" objects, not only instance variables pointing to +neighbours) tmu: what about creating a new key, which is not a byte object but +contains a byte object tmu: the object would return the key object, and the key would also be +the 'key object' tmu: Changing the key, however, would not exchange the key object, but +only the byte object which is the contents of the key object *** Signoff: tonyKwong (ccwf.cc.utexas.edu penfold.ece.uiuc.edu) Then I loose the performance advantages which are the reason for this +redundant representation, you could then as well take the objects as keys AND +values, which is the same as using the least structured collection: a set *** tonyKwong has joined channel #programST *** Mode change "+o tonyKwong " on channel #programST by penfold.ece.uiuc.edu tmu: Even if you store and access the new key objects only? tmu: The key objects are still unique objects! The problem lies in another design flaw of Smalltalk. Most interface +objects like meus or lists force you to convert objects to strings (for +display) and vice versae after you had your choice. Wth the string returned +you have to look up the corresponding objects tmu: I had a similar problem with listpanes tmu: I had a list of objects , generated a list of strings to display +them. I have posted some goodies: ObjectListPanes and ObjectMenus which work +with REAL objects and a block to tell the list/menu how the objects have to +be displayed. But you get real objects when you click on them. After that, however, I didn't search the list of objects for the +selected one, but took the selection index! to access it That sounds similar to the problem I had: You cant rely on object +identity, one of the MAIN characteristics of OO technology! :-) My version of ObjectListPane is called NewListPane funny to see different people solving the same problem in a similar way There are similar problems There's another kind of pane if would propose to consider The good thing with OO lists and meus is that you can use identical +item-displays with different objects "behind" them. With the common method +you scanned your collection via "detect:" and always got the first occurence +of a corresponding objec rather often I needed a kind of list pane that didn't show only one +selection, but, independently from the main selection, a group selection Not only a MultipleChoiceList? (multiple selecte list pane ?) Think of lists of files or other things. There is one file you want to +see or to open, but there may be some files to delete or copy at a time Not quite A normal multiple select list pane has only one kind of selection you may select more than one item, but all of them are equally selected What I needed, was a pane with two different kinds of selection: a +group selection and a single selection which would be represented in a +different way at the same time or as alternates? After I created such a window for my project environment, I found it +quite usefull for other applications. At the same time How did you manage not to make the procedure at the interface too +cumbersome? to give an example: You may select a class in my ClassHierarchyBrowser quite normally. The +methods of this class will be shown as usual If you hold the shift key, you may select a number of classes. The CHB +doesn't display the methods of these classes, but you may (e.g. from a menu) +decide to print, file out or remove them while you're inspecting the methods +of your maPRIVMSG #programst :No The point is, that sometimes you have rather different actions on the +same class of objects actions, which are usefull for a group of objects *** smith has joined channel #programST and actions which are usefull for a single object. welcome smith Thanks Rolf: I would consider this procedure ok for developers but I doubt that +this will be a consistent approach for the design of "normal" user interfaces +- at least it is not CUA!-) inspecting the methods of a class is an action that needs the user to +look at the CHB and the CHB time to show the methods Hello smith When I want to print some classes I don't want the CHB to show me the +methods of each class when it select it for printing That's why I use 2 differnt kinds of selection tmu: No, its no CUA But the people at my group got rather fast accustomed to it *** Signoff: smith (Ping timeout) *** smith has joined channel #programST If somebody is interested in beta-testing our current implementation of +a "text interpretation support system", you can send a mail to +ATLAS@db0tui11.bitnet. The system is written in Smalltalk/V 286, needs a 386 +PC with at least! 4MB and a VGA card. A geman and english version (with manual) are available. Dreaded writing bugs! :) Could that be interesting for people doing anglistics or germanistiks +or something like that ? We already have it tested by people from the social sciences, +linguistics, knowledge aquisition literature science, theology, tec. All areas where TEXT is the central +material and has to be interpreted. But: the program does not interpret anything, it only offers tools for +the researcher. tmu: if you need betatesting, I could ask a friend of mine doing +germanistics. However, I don't know what kind of PC he's using There is also a theory building component (creating hypertext and +semantic networks) with the option for converting the structures into a +Prolog format. You can also create SPSS-jobs if you want to do quantitative analysis as +well.