On 28/Mar/2014, at 11:52 , Christiaan Hofman wrote:
On Mar 28, 2014, at 4:56, Fischlin Andreas wrote: Hi all, Despite what Christiaan said, I insist that what I suggested would be a good solution. Christiaan seems still not to understand what I suggested. I do understand. I just see it differently, and most definitely the implementation conflicts with this way of seeing it. A clear user interface is generally helping the implementation and the implementation should never dominate the UI. As I explained earlier, the user model is what counts at the UI level. The implementation has to follow that. This means, in the user model of BibDesk we have "Search Group Settings" (SGS) that are identified by a name. The name does not really matter much except that it should the user help to quickly identify the SGS. Each SGS has associated a set of parameters. One is the server and then some server settings. The web services we are talking about here are for WOS or WOK. (The mess Christiaan is mentioning we have "thanks" to Thomson Reuters, who can never decide what they want to call their services. I suggest to use WOK, since from all what I know, this is the most appropriate term to use since about a year.) But their web pages seems to say Web of Science, not Web of Knowledge, even if their domain name is webofknowledge.com<http://webofknowledge.com/>. And is it WoK, WoS, or WOK, WOS? You are right it is a mess, as I said. My take on all the discussions I had with them on this is WOK. WOS is rather one server only. Yes, it shows in the title of the web page by default "WOS". This is because this is the default data base to use. But WOS is only the name of one of the many data bases, that are offered by the WOK services. Therefore the server's URL is http://apps.webofknowledge.com. Other data bases are BIOSIS, Current Contents, SCI etc. The name Thomson Reuters uses for the web service we are accessing through BibDesk is WOKSearch or then WOKSearchLite, depending on subscription (i.e. money you pay them). WOKSearch is also called WOKSearchPremium. Thus I believe we best use in BibDesk now 'WOK' as a short name of WOKSearchPremium services, through which we can access WOS, BIOSIS, SCI etc. The not premium access to those services I suggest we best call 'WOK Lite' accordingly. BTW, AFAIK again the number of data bases actually available through WOKSearchPremium or WOKSearchLite depend again on the subscription. BibDesk comes with some predefined SGS, one of them giving access to WOK searches authenticating via IP#. Any "Server Settings" associated with a SGS accessing these so-called web services from Thomson Reuters are currently only accessible by using the server setting 'ISI', which corresponds to what Thomson Reuters calls the 'WOKSearchPremium' web service or a bit more abbreviated 'WOKSearch'. The lite variant of that is called by Thomson Reuters 'WOKSearchLite'. From the user model there is nothing more to be added and I can not see why any implementation issue would interfere with this user model. But it does. The radio button selects what type of server to use. And the premium and lite version use the same type of server in its implementation. Whether it then uses the premium or lite version is a choice that depends on an option. Yes, I know. To be frank, I believe you are looking too much at these things from the implementer's view. I concur with you, that your argumentation "naturally" can be seen as calling for simply adding a checkbox. But then you forget how that looks to every user. In particular given you argue that there is no space to put that checkbox near the radio button to which it belongs. Then you really open up the use of this checkbox with any other than the ISI radio button too. Why should a user then understand that this option matters only for one particular radio button choice, but not for others? The radio button solution I suggest, however, is in this respect much cleaner. It implements the actual user model: You use the 'WOK' or the 'WOK Lite' service from Thomson Reuters. Period. If you want to add a checkbox far away from the radio buttons, then I ask you to at least to make it inactive unless the radio button choice 'ISI' (or newly labelled 'WOK') is currently in use. That might help the user to understand the relationship. In both cases you're choosing using the WoK server. Then it is a choice as to how it connects to the server, or if oyu like what type of search you do., I.e. through the premium or lite channel. That's an option of the WoK search group, not a type of search group, as I see it. The dialog to view or edit SGS's we are talking about here should fully support this user model. Since you can only make WOK searches either using the WOKSearchPremium or the WOKSearchLite service, but never both at the same time, radio button choice is the best thing you can do in any GUI (exclusive or). A checkbox (boolean) is messy, since you can use that checkbox in combination with any other radio button choice. However, as I showed above, any other combination than with the radio button 'ISI' makes no sense. If you have a checkbox you confuse only the user model, since in general the use of a checkbox together with a set of radio buttons implies always you can make any combination. That's false. The check box is only there for the ISI group. Just as the address field is only there for the z39.50 group. Oh, I see. Ok. I did not understand your dialog picture that way. Showing the checkbox only when the ISI (WOK) server setting is in use, that is even better than making the checkbox inactive vs. active. Then I agree that this checkbox solution is also an option. But only under this condition. Finally, it is a little detail how to label the radio buttons. I tried to stick to previous labels, i.e. 'ISI' and added accordingly 'ISI Lite'. I admit that can be questioned and one could argue, it might be better to rename radio button 'ISI' to 'WOK' and have then as the new radio button 'WOK Lite'. I could live with both solutions. Therefore only the radio button choices I suggested make fully sense and are really convincing. I also see not why there should follow from this dialog design any implementation problems, since the dialog is factually modal. I can see that only when you mix implementational details with a GUI design and let the implementation spill into the user model and make that visible at the GUI level. This is never a good idea and only causes user troubles. Good programing can and should IMHO avoid that. Cheers, Andreas You see premium and lite as two basically completely unrelated types of search servers. I see it as the same type of search servers, but with an extra option telling it how to search. Just like for Zoom you have to say which server address to use, here you have to say whether to use the premium or lite channel. So it's a subchoice. I don think that makes sense for most users, they know they want to search WOK, they won't see WOK and WOK Lite as two different types of search servers. Quite frankly, I think WOK and WOK Lite are much more the same than, say, Library of Congress and Harvard. The latter are the same type (zoom), wo why not the former? I think that absolutely makes sense. I agree there is also merit to this view. Some of these search services (note, I do not say servers on purpose) are more related than others. I would prefer the use of the word service, not server. You may well use the same server as given by a particular URL for different services or the same service via different servers (AFAIK actually the case with WOK). The different choices for the radio buttons describe completely different protocols for the servers, i.e. completely different types of servers. That is also directly associated with the implementation: we use completely different server classes for the different types. The WOK servers are basically the same protocol, they just use some different settings (like the address.) They (will) also use the same server class in the implementation. This argument is the one I do not accept. Good software always differentiates between the user model and the implementation model. They do rarely go together and actually should not go together. The user model should be as simple as possible, which serves the purpose to be easy to understand. Therefore always an abstraction from the implementation and you should not require the user to understand the typically much more complex implementation in order to be able to use the program. Finally, OOP is no excuse for not making and seeing the difference between the user model and the implementation, although I admit it seduces to do so. Cheers, Andreas Christiaan On 27/Mar/2014, at 22:43 , Christiaan Hofman wrote: On Mar 27, 2014, at 16:56, Reto Stöckli wrote: How about using the Database field in the search group settings? one can enter either nothing, Premium or Lite? Which is really what it does. Changing Database (subsetting) and changing output style. If you use the WOKSearch Premium, you access a bigger set of the database than when you search WOKSearch Lite. It's not just the number of output fields, it is also the number of retrievals per search. Reto I am a bit hesitant with this. It is probably the quickest and easiest to implement, but it does not really feel like a database setting (i.e. a choice of database). So this would be very counter intuitive. And then there's the conflict with the old way, so backwards compatibility problems. Christiaan ------------------------------------------------------------------------------ _______________________________________________ Bibdesk-develop mailing list [email protected]<mailto:[email protected]> https://lists.sourceforge.net/lists/listinfo/bibdesk-develop ------------------------------------------------------------------------------ _______________________________________________ Bibdesk-develop mailing list [email protected]<mailto:[email protected]> https://lists.sourceforge.net/lists/listinfo/bibdesk-develop Christiaan ------------------------------------------------------------------------------ _______________________________________________ Bibdesk-develop mailing list [email protected]<mailto:[email protected]> https://lists.sourceforge.net/lists/listinfo/bibdesk-develop ------------------------------------------------------------------------------ _______________________________________________ Bibdesk-develop mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/bibdesk-develop
