On Thu, 24 Oct 2013, Talha Yalta wrote: There was a long debate between March and May on the respective merits of the two choices, and IMO we collectively took a very balanced decision. You may want to start from http://lists.wfu.edu/pipermail/gretl-devel/2013-April/004398.html I don't think we should revisit the decision we made at that time. Besides, bear in mind that one instance of Openoffice enables you to work on more than one document at a time, whereas you can't work on more than one dataset at a time from one gretl instance, so if you want/need to have two dataset open simultaneously, you've got to have multiple instances. Comparison with R doesn't really apply IMO, because in R there's no such a thing like the "current dataset" we have in gretl. This makes sense, I agree. ------------------------------------------------------- Riccardo (Jack) Lucchetti Dipartimento di Scienze Economiche e Sociali (DiSES) Università Politecnica delle Marche (formerly known as Università di Ancona) r.lucchetti(a)univpm.it http://www2.econ.univpm.it/servizi/hpp/lucchetti -------------------------------------------------------

I am not affected with this multiple instances thing, but if possible a Preferences entry like: Ask to open a new instance (default) [x] (disabled in case of default active; Radio buttons) Open a new instance: Always / Never On Thu, Oct 24, 2013 at 9:08 AM, Talha Yalta <talhayalta(a)gmail.com&gt; wrote:

On Thu, 24 Oct 2013, Riccardo (Jack) Lucchetti wrote: I agree we shouldn't reopen that question right now, just before a release. However, we may want to consider refining the "multiple instances" business later. For the record, here's a summary of the situation as I see it: 1) Until we introduced the "Start a new instance?" query, users on Windows and X11 always ended up starting a second instance of gretl whenever (a) gretl was already running and (b) they did something that normally starts the program (selecting gretl off a desktop menu, double-clicking a gretl program icon, double-clicking a gretl-associated file). That is, gretl had no mechanism for enforcing uniqueness of the running instance. 2) On OS X, on the other hand, the OS itself enforces uniqueness, so you'd get different behaviour. And on Windows and X11, some programs have their own uniqueness-enforcement mechanism (e.g. using libunique or the GApplication API). 3) As Jack says, totally enforcing uniqueness in gretl wouldn't be appropriate, since you can only have one dataset open in a given gretl instance. Some users, some of the time, may really wish to have more than one instance running, so we give a choice. 4) However, as Talha implies, most users most of the time won't want to run more than one instance, and faced with the question "Start another instance?" they may think, "Of course not, why would you ask?" If we were going to modify this, I think two ideas may be worth considering: 1) Switch gretl on Windows and X11 to the same mechanism we have on OS X. That is, enforce uniqueness in general, but provide a specific menu item to start another instance, overriding the uniqueness check. 2) Keep our current "Start another instance?" dialog, but add a check box with text something like "Never start another instance": if you check that, you won't see the dialog again. This could be linked to a preferences item, "Allow multiple instances". So you could get the dialog back if you checked that. (My idea is that this preference item would be checked by default, but would become unchecked if you took the "Never" option in the query dialog.) I don't think it's crucial that we address this before release, since most of the time users won't see that query dialog unless they lose track of a running instance of gretl, in which case it doesn't seem like a lot of trouble to click "No". Allin