Hi Maarten, *, sorry for the delay, this sat in my outbox unnoticed..
On Nov 26, 2007 9:30 PM, :murb: [maarten brouwers] <[EMAIL PROTECTED]> wrote: > > Good questions, and you may be right that a one click is not the best > currently. But let us consider one click for now, it would be easiest if > it would work perfect. I assume, and most do, that you download OOo for > the system you are working on. Only power users like you, and me, may > download OOo for somebody else, burn it on CD and give it away... but I > don't think that this is most typical behaviour (we might try to get > some data on this by monitoring the current contribution page... While it is less the case with small things like firefox, OOo is quite a bit to download, so here the case where some person downloads OOo with a fast connection for somebody else with a slower connection is more common IMHO. But that aside: My concerns are not about whether it is technically possible to sniff the operating system or the language and even whether a java-plugin is active in the browser. My concern is: Do these values make sense at all? Many browsers are not set to the language the user wants to get, many users don't have java activated because of performance reasons or security concerns. (even when java is enabled, would you still suggest the JRE included version when the bundled JRE is newer than the one on the users' machine? JRE is a big download to an already big OOo. JRE is an optional thing, many users will never use it, they will use OOo as a typewriter that can do underlined, bold and italics... But there is more, see below. > > OOo has versions in RPM and DEB format for Linux (and some with JRE, > > some without) > > Can we discern between Debian deriviates (among which ubuntu) and > redhats (thats rpm right?) using the HTTP_USER_AGENT string? See, this is another hack.. Of course you can tell, firefox is relabled iceweasel on Debian (or whatever), but again: Does a detection make sense? > [...] > > Default to english and still start the download? Bad, bad, bad for > > non-english users... > > There is still one click required to start it... Sure - but on the one hand you demand the native-lang projects to be equally represented on all the pages, and then you "only offer an english download" (yes, there would be a link next to the big button, but many folks will overlook that for sure) > > Default to version with JRE? Bad, bad, bad when the nl-community is > > not QA'ing the version of OOo that has JRE included, doesn't build > > versions with JRE at all (AFAIK only Sun does build versions with JRE) > > Does it matter technically? It matters in terms of which installsets end up on the mirror-network. > I mean, do you think that a QA for an without JRE version differs from a > QA with JRE, or are you talking about availability? I'm talking about availability - and of course there might be installer bugs that only happen in the with-JRE version, but again this is only one of the issues with a preselection. User without Java plugin: Probably the user has a reason for not using java. Your reasoning now is: The user is missing java, so he should get the installset with JRE included. The users is bothered because he needs to switch to some other download page (that very likely will not get the same level of care than the main download page...). User with java-plugin: Is using java, so he might want to use java with OOo OOo's java might be newer, OTOH some other legacy applications (critical, financial stuff) only works with older versions. Installing a new JRE will break that application for the user, leaving the user clueless about how to fix the problem. (Not a theoretical problem, saw that in reality already, of course the vendor's tech-support is clueless as a log of wood, and vendor is not willing to release an updated version that is upwards-compatible). > > Native-language versions in general: How would you deal with different > > versions being QA'd / approved at different points in time? > > Report version in the button So you rely on the user to know what version is the latest one? The user downloads the "old" version only to get notified two days later via the online-update that a new version is available, so that the user will have to download OOo twice? :-/ > >> Costs: somebody needs to do the javascript needed for this. If we focus > >> on only OS, this is very easy. > > > > No, it is not. We have multiple versions per OS, and even more when > > you get to the languages. > > Well, it is not hard when we are not considering only English... I guess there is a "not" too much in that statement. And still I disagree that it would be easy. Don't get me wrong, the script to get the values for OS, language and java might be easy to retrieve via javascript, but still this won't be enough to make a sensible default. It is easy for firefox, there you have english and "OS" - no additional consideration of JRE... And it boils down to: (and I apologize to Kay for the wording - in germany every male term needs to be "feminized" Schüler gets Schülerin (or SchülerIn with uppercase I) to put females on an equal level - of course that backfired in english where "user" is always "user", whether female or not. "Benutzer" OTOH is male, Benutzerin would be the female one). > > I /still/ don't see a one-click solution, and honestly I still don't > > see what is wrong when > > the user has to do two clicks. Is the user really that dumb so that > > she doesn't know what > > operating system she runs? > > Some believe/know(?) users are really that dumb. I am not sure, but I do > know that every additional cognitive effort required for something users > are not sure about is another point where we can loose this person. > Since we want as many users to download OOo as possible, it important to > reduce chances in loosing users (can we have data on this?) Well, I have a slightly different goal as well. For example I completely disagree with Louis vision to have a one-click download that will hide the other download options such like p2p/bittorrent. But even if I had the same wish: I still don't believe that a default selection will be what the user wants. I at most would go with detection of the Operating system and then only present the options for that OS (for Windows it would be two buttons/choices: with JRE, without JRE. For mac it would be multiple choices at once (PPC vs Intel, Panther vs tiger) or in that case you probably can rely on getting the right OS-architecture. For linux it would present the deb/rpm and with and without jre options, ..) I would /not/ automatically detect the JRE choice for the reasons mentioned above: I believe there cannot be a good guess based on activated java-plugin in your browser. And I would /not/ automatically guess the langauge for the reasons mentinoed above (you cannot know what versions are already QA'd, distributed on the mirrors, and you cannot rely on the browser's language either) > > I fear that all of those "downloading is too complicated" boils down > > to this very additional > > click on the contributing page.... > > I definitely think that this is indeed one of the main concerns... and > the delayed download with a contribution page is preferred by me as well. > > We can offer one click download for only those who we can reliably > detect, or offer the right download option with reasonably certainty > (i.e. IE users -> Windows download, and just add JRE to it) I still disagree with what is a reliable detection. I don't question that you can detect whether the user is running on windows or on some other system and whether he has java activated in the browser or not. I guestion that you can choose a default installation package based on that data. ciao Christian PS: When it comes to implementing a javascript based check - you don't have to reinvent the wheel, google just gives a bunch of hits.... http://www.webreference.com/tools/browser/javascript.html basically already does all the stuff - license: GPL (view page source) http://techpatterns.com/downloads/javascript_browser_detection.php another one (GPL as well) --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
