On Wed, Apr 13, 2016 at 11:15 AM, Guillaume Delhumeau < [email protected]> wrote:
> I don't like the option 2 because it mentions "space" which is not > presented in the UI and which will disappear eventually. > > On the other hand, option 3 has the drawback of writing "WebHome" > everywhere, which is again something we want to hide. > > Since we have recently decided that "page" should be the word used instead > of "document", and since the creation of "terminal page" is only an option > for advanced users, I propose to introduce the type "page" as an alias of > "space". > > So to link a terminal page, we would use "doc:A.B" and to link a nested > page, we would use "page:A.B" (which will be resolved as A.B.WebHome). No > ambiguity. > > With that proposition, I would be for the option 2 > > I'm sorry to hijack the vote with a new proposition, but I though it was > important. > I was actually thinking of the same thing. Maybe we should have chosen "page" from the start. If we move to "page" I would prefer to replace "space" instead of adding yet another alias for it (which would not be as trivial as it seems) since we`re still rather early in the Nested Page adoption. We could probably try some migrator as well for that. | Regarding the link syntax complexity, IMO the source of all these issues is the "default" and "indirect"/"accidental" relativity of [[A]] links. I say "indirect"/"accidental" because it`s not designed, IMO, but only a side effect of the way the reference resolver works and, together with the "relative" effect, we also get all these weird limitations that don`t make sense for a user wanting page references and not document references (witch also have a space element). A solution to this, IMO, would be to make references absolute by default and have relative links be explicit (probably with the [[.A]] notation and the [[..A]] notation for parent, as Vincent was saying). Migrations would be problematic though. Another solution (maybe more manageable in terms of migrations, and more user friendly) would be a new link syntax, as Vincent again mentioned, but then we would find ourselves translating the link syntax to the reference syntax which would now diverge (even more than they did in order to be able to avoid ".WebHome"). | Regarding the actual issue at hand, I agree with Marius on the 3rd option, since that is what I have also implemented in the WYSIWYG (in order to accept the new "space:" link type) and it does make sense from a visual editor point of view. If users use a visual editor they should not care about what it produces. Think of the old DreamWeaver solution or any other visual tool that produces code... it will never be pretty code but it will work and you can never expect it to be pretty enough, as if a human wrote it. It needs to cover many use cases as best as possible, whereas if a human wrote it, it would cover a single specific use case very neatly. If some users use WYSIWYG and others use wiki syntax there will surely be some friction by design and I don`t see how we can get around it. Thanks, Eduard WDYT? > > Thanks, > > > > 2016-04-12 16:29 GMT+02:00 Vincent Massol <[email protected]>: > > > > > > On 12 Apr 2016, at 16:18, Marius Dumitru Florea < > > [email protected]> wrote: > > > > > > On Tue, Apr 12, 2016 at 5:12 PM, Thomas Mortagne < > > [email protected]> > > > wrote: > > > > > >> On Tue, Apr 12, 2016 at 4:09 PM, Vincent Massol <[email protected]> > > >> wrote: > > >>> > > >>>> On 12 Apr 2016, at 15:43, Marius Dumitru Florea < > > >> [email protected]> wrote: > > >>>> > > >>>> On Tue, Apr 12, 2016 at 4:30 PM, Vincent Massol <[email protected] > > > > >> wrote: > > >>>> > > >>>>> Hi Marius, > > >>>>> > > >>>>>> On 12 Apr 2016, at 14:56, Marius Dumitru Florea < > > >>>>> [email protected]> wrote: > > >>>>>> > > >>>>>> Hi devs, > > >>>>>> > > >>>>>> We need to decide what is the expected link reference > serialization > > >>>>>> produced by the WYSIWYG editor when you create a link to a wiki > > page. > > >> I > > >>>>>> think we have to choose between 3 options: > > >>>>>> > > >>>>>> (1) Output untyped (ambiguous) link references whenever possible > > >>>>>> (2) Always output unambiguous (typed) link references (both 'doc:' > > and > > >>>>>> 'space:') > > >>>>>> (3) Always output 'doc:' link references (no 'space:' references) > > >>>>>> > > >>>>>> Let's see the details: > > >>>>>> > > >>>>>> = OPTION 1: Output untyped link references whenever possible = > > >>>>>> > > >>>>>> == <7.2 == > > >>>>>> > > >>>>>> Link from A.B > > >>>>>> * to A.WebHome (space home page) => [[WebHome]] > > >>>>>> * to A.C (same space) => [[C]] > > >>>>> > > >>>>> I don’t understand why you’re using relative links in your 2 > examples > > >>>>> above. Option (1) says untyped, it doesn’t say convert links into > > >> relative > > >>>>> links. > > >>>>> > > >>>>> So for me that would be: > > >>>>> > > >>>>> * to A.WebHome (space home page) => [[A.WebHome]] > > >>>>> * to A.C (same space) => [[A.C]] > > >>>>> > > >>>>>> * to X.Y (different space) => [[X.Y]] > > >>>>>> * to X.WebHome => [[X.WebHome]] > > >>>>>> > > >>>>>> == 7.2+ == > > >>>>>> > > >>>>>> Link from A.B.WebHome > > >>>>>> * to A.WebHome (its parent) => [[space:A]] > > >>>>> > > >>>>> Note 1: We could also imagine introducing a syntax for absolute > links > > >> such > > >>>>> as [[:A]] > > >>>>> Note 2: We could also imagine introducing a syntax for parent links > > >> such > > >>>>> as [[..A]] > > >>>>> > > >>>>>> * to A.B.C or A.B.C.WebHome or A.C or A.C.WebHome (child or > sibling) > > >> => > > >>>>>> [[C]] (very ambiguous) > > >>>>> > > >>>>> This is using a relative notation. But (1) is not about > transforming > > >> links > > >>>>> into relative links. > > >>>>> > > >>>>> so for me this is: > > >>>>> * => [[A.B.C]] > > >>>>> * => [[A.C]] > > >>>>> > > >>>>> I’m stopping reading here since first we need to clarify if (1) is > > >> about > > >>>>> a) untyped or b) using relative references. > > >>>>> > > >>>>> IMO it should be about a) and doesn’t have to do b). > > >>>>> > > >>>>> > > >>>> > > >>>>> Maybe I misunderstood something? > > >>>>> > > >>>> > > >>>> Generating relative references is a must for me (whenever possible), > > no > > >>>> matter what output format we choose. I'm surprised there's any doubt > > >> about > > >>>> this. The current WYSIWYG editor is generating relative references > and > > >> the > > >>>> CKEditor should continue to do this. Relative references have many > > >>>> advantages (eases the refactoring, allows you to export a hierarchy > > and > > >>>> import it somewhere else, etc.). > > >>> > > >>> Didn’t realize the WYSIWYG was already doing this! > > >>> > > >>> Indeed there are pros. The cons is that it can make the linking more > > >> ambiguous., but overall it’s probably a good thing. > > >>> > > >> > > >>> Also note that ideally we’d need to not convert existing links when > > >> going through the WYSIWYG editor. That’s true independently of this > > >> discussion though. > > >> > > >> > > > > > >> This is handled by XML comments containing the source reference so the > > >> only thing the WYSIWYG need to do is to not touch those comments > > >> (unless the user explicitly modify the target of course). > > >> > > > > > > It's already the case. The link reference is preserved as is unless you > > > edit the link and apply the changes (i.e. you close the link dialog > with > > > the OK button). > > > > Note that the majority of users will edit a link and click ok even if > they > > don’t change anything. An improvement would be to check for a difference > in > > the generated link. > > > > What I was commenting on though was that if the user has chosen an > > absolute format or a relative format and he changes some link-related > data > > (label, target location, etc), ideally the WYSIWYG would output a > reference > > using the same format (absolute or relative). > > > > Thanks > > -Vincent > > > > > Thanks for the reply, I understand now. > > >>> > > >>>> So option (1) is about untyped relative links, option (2) is about > > typed > > >>>> relative links and option (3) is about doc: relative links. In other > > >> words: > > >>>> (1) don't generate "doc:" and "space:" > > >>>> (2) generate both "doc:" and "space:" > > >>>> (3) generate only "doc:" (don't hide WebHome) > > >>> > > >>> I’ve now read again the first mail and I’m in favor of (2). > > >>> > > >>> The main reason for me is that we want to hide WebHome and with > Nested > > >> Pages, all links would have WebHome in them with option (3). > > >>> > > >>> Option (3) is really the worst for users: they get to see both “doc:” > > >> and “WebHome” :) > > >>> > > >>> Also note that even option (2) is not perfect because of the “doc:” > and > > >> “space:” prefixes which are also hard to understand for users. > > >>> > > >>> I also know users who use both WYSIWYG and wiki editors (and who > think > > >> that WebHome is confusing - We’re used to it, but it’s really > confusing > > if > > >> you’re just starting to use XWiki - you don’t even understand what it > > means > > >> at all). > > >>> > > >>> To summarize: +1 for (2). > > >>> > > >>> Note: My choice for (2) is based purely on a usability POV. It could > be > > >> that implementing option (2) is so much complex that it’s not worth it > > and > > >> that we’d want to wait till we change the underlying model before > > changing > > >> the linking syntax. > > >>> > > >>> Last note: Maybe we need to invent some completely new syntax for > > >> linking since currently it has become very complex. I remember of any > > mail > > >> where I proposed some new syntax for links using a different syntax > > such as > > >> [[[…]]]. It could be interesting to spec a new simpler syntax such as: > > >>> * Use “/“ instead of “.” (we already know that users would prefer “/“ > > >> since they’re more used to that symbol) > > >>> * Leading “/“ means absolute. Example: [[[/A/B]]] > > >>> * Use “..” for parent. Example: [[[..A/B]]] > > >>> > > >>> Thanks > > >>> -Vincent > > >>> > > >>>> Thanks, > > >>>> Marius > > >>>> > > >>>> > > >>>>> > > >>>>> Thanks > > >>>>> -Vincent > > >>>>> > > >>>>>> * to A.B.C.D or A.B.C.D.WebHome (nested page two levels or more > > >> below) => > > >>>>>> [[.C.D]] (ambiguous) > > >>>>>> * to A.C.D or A.C.D.WebHome (nested page under sibling) => > [[A.C.D]] > > >>>>>> (ambiguous) > > >>>>>> * to X.Y.Z or X.Y.Z.WebHome => [[X.Y.Z]] (ambiguous) > > >>>>>> > > >>>>>> Link from A.B (terminal) > > >>>>>> * to A.WebHome (its parent) => [[space:A]] > > >>>>>> * to A.C or A.C.WebHome (sibling) => [[C]] (ambiguous) > > >>>>>> * to A.C.D or A.C.D.WebHome (nested page under sibling) => > [[A.C.D]] > > >>>>>> (ambiguous) > > >>>>>> * to X.Y or X.Y.WebHome => [[X.Y]] (ambiguous) > > >>>>>> > > >>>>>> Link from A.B.C (terminal) or A.B.C.WebHome > > >>>>>> * to A.B.WebHome (its parent) => [[A.B]] (ambiguous) > > >>>>>> > > >>>>>> PROS: > > >>>>>> * shorter link references > > >>>>>> * hides WebHome from source syntax on 7.2+ > > >>>>>> > > >>>>>> CONS: > > >>>>>> * ambiguous link references > > >>>>>> * complex code > > >>>>>> * different output for <7.2 and 7.2+ in case of top level space > > >>>>> [[WebHome]] > > >>>>>> or [[A.WebHome]] vs. [[space:A]] (we need to check if support for > > >> nested > > >>>>>> spaces is available) > > >>>>>> > > >>>>>> > > >>>>>> = OPTION 2: Always output unambiguous link references = > > >>>>>> > > >>>>>> == <7.2 == > > >>>>>> > > >>>>>> Link from A.B > > >>>>>> * to A.WebHome (space home page) => [[doc:WebHome]] (relative) > > >>>>>> * to A.C (same space) => [[doc:C]] (relative) > > >>>>>> * to X.Y (different space) => [[doc:X.Y]] > > >>>>>> * to X.WebHome => [[doc:X.WebHome]] > > >>>>>> > > >>>>>> == 7.2+ == > > >>>>>> > > >>>>>> Link from A.B.WebHome > > >>>>>> * to A.WebHome (its parent) => [[space:A]] > > >>>>>> * to A.B.C (terminal child) => [[doc:C]] (relative) > > >>>>>> * to A.B.C.WebHome (non-terminal child) => [[space:A.B.C]] > > (absolute) > > >>>>>> * to A.C (terminal sibling) => [[doc:A.C]] > > >>>>>> * to A.C.WebHome (non-terminal sibling) => [[space:A.C]] > > >>>>>> * to A.B.C.D (terminal descendant) => [[doc:.C.D]] (relative) > > >>>>>> * to A.B.C.D.WebHome (non-terminal descendant) => > [[space:A.B.C.D]] > > >>>>>> (absolute) > > >>>>>> * to A.C.D (terminal descendant of sibling) => [[doc:A.C.D]] > > >>>>>> * to A.C.D.WebHome (non-terminal descendant of sibling) => > > >>>>> [[space:A.C.D]] > > >>>>>> * to X.Y.Z => [[doc:X.Y.Z]] > > >>>>>> * to X.Y.Z.WebHome => [[space:X.Y.Z]] > > >>>>>> > > >>>>>> Link from A.B (terminal) > > >>>>>> * to A.WebHome (its parent) => [[space:A]] > > >>>>>> * to A.C (terminal sibling) => [[doc:C]] (relative) > > >>>>>> * to A.C.WebHome (non-terminal sibling) => [[space:A.C]] > (absolute) > > >>>>>> * to A.C.D (terminal descendant of sibling) => [[doc:.C.D]] > > (relative) > > >>>>>> * to A.C.D.WebHome (non-terminal descendant of sibling) => > > >>>>> [[space:A.C.D]] > > >>>>>> * to X.Y => [[doc:X.Y]] > > >>>>>> * to X.Y.WebHome => [[space:X.Y]] > > >>>>>> > > >>>>>> Link from A.B.C (terminal) or A.B.C.WebHome > > >>>>>> * to A.B.WebHome (its parent) => [[space:A.B]] (absolute) > > >>>>>> > > >>>>>> PROS: > > >>>>>> * unambiguous link references ("what you link is what you get") > > >>>>>> * slightly less complex code (but still complex) > > >>>>>> * hides WebHome from source syntax on 7.2+ > > >>>>>> > > >>>>>> CONS: > > >>>>>> * longer link references (because of "doc:" and "space:" prefixes) > > >>>>>> * cannot specify relative 'space:' references > > >>>>>> * different output for <7.2 and 7.2+ in case of [[doc:WebHome]] > vs. > > >>>>>> [[space:A]] > > >>>>>> > > >>>>>> > > >>>>>> = OPTION 3: Always output 'doc:' references = > > >>>>>> > > >>>>>> == <7.2 == > > >>>>>> > > >>>>>> Link from A.B > > >>>>>> * to A.WebHome (space home page) => [[doc:WebHome]] (relative) > > >>>>>> * to A.C (same space) => [[doc:C]] (relative) > > >>>>>> * to X.Y (different space) => [[doc:X.Y]] > > >>>>>> * to X.WebHome => [[doc:X.WebHome]] > > >>>>>> > > >>>>>> == 7.2+ == > > >>>>>> > > >>>>>> Link from A.B.WebHome > > >>>>>> * to A.WebHome (its parent) => [[doc:A.WebHome]] > > >>>>>> * to A.B.C (terminal child) => [[doc:C]] (relative) > > >>>>>> * to A.B.C.WebHome (non-terminal child) => [[doc:.C.WebHome]] > > >> (relative) > > >>>>>> * to A.C (terminal sibling) => [[doc:A.C]] > > >>>>>> * to A.C.WebHome (non-terminal sibling) => [[doc:A.C.WebHome]] > > >>>>>> * to A.B.C.D (terminal descendant) => [[doc:.C.D]] (relative) > > >>>>>> * to A.B.C.D.WebHome (non-terminal descendant) => > > [[doc:.C.D.WebHome]] > > >>>>>> (relative) > > >>>>>> * to A.C.D (terminal descendant of sibling) => [[doc:A.C.D]] > > >>>>>> * to A.C.D.WebHome (non-terminal descendant of sibling) => > > >>>>>> [[doc:A.C.D.WebHome]] > > >>>>>> * to X.Y.Z => [[doc:X.Y.Z]] > > >>>>>> * to X.Y.Z.WebHome => [[doc:X.Y.Z.WebHome]] > > >>>>>> > > >>>>>> Link from A.B (terminal) > > >>>>>> * to A.WebHome (its parent) => [[doc:WebHome]] (relative) > > >>>>>> * to A.C (terminal sibling) => [[doc:C]] (relative) > > >>>>>> * to A.C.WebHome (non-terminal sibling) => [[doc:.C.WebHome]] > > >> (relative) > > >>>>>> * to A.C.D (terminal descendant of sibling) => [[doc:.C.D]] > > (relative) > > >>>>>> * to A.C.D.WebHome (non-terminal descendant of sibling) => > > >>>>>> [[doc:.C.D.WebHome]] (relative) > > >>>>>> * to X.Y => [[doc:X.Y]] > > >>>>>> * to X.Y.WebHome => [[doc:X.Y.WebHome]] > > >>>>>> > > >>>>>> Link from A.B.C (terminal) > > >>>>>> * to A.B.WebHome (its parent) => [[doc:WebHome]] (relative) > > >>>>>> > > >>>>>> Link from A.B.C.WebHome > > >>>>>> * to A.B.WebHome (its parent) => [[doc:A.B.WebHome]] (absolute) > > >>>>>> > > >>>>>> PROS: > > >>>>>> * unambiguous link references ("what you link is what you get") > > >>>>>> * relative references for nested non-terminal descendants > > >>>>>> * same output for <7.2 and 7.2+ > > >>>>>> * simpler code (easier to maintain) > > >>>>>> > > >>>>>> CONS: > > >>>>>> * doesn't hide WebHome from source syntax > > >>>>>> * longer link references (because of "doc:" prefix and "WebHome" > > >> suffix) > > >>>>>> > > >>>>>> > > >>>>>> I don't like ambiguous links so I'm against option (1). I don't > > think > > >> the > > >>>>>> WYSIWYG editor users care too much about the wiki syntax (option > 2), > > >> as I > > >>>>>> commented on http://jira.xwiki.org/browse/XWIKI-13083, so I'm +1 > > for > > >>>>> option > > >>>>>> (3). > > >>>>>> > > >>>>>> WDYT? > > >>>>>> > > >>>>>> Thanks, > > >>>>>> Marius > > _______________________________________________ > > devs mailing list > > [email protected] > > http://lists.xwiki.org/mailman/listinfo/devs > > > > > > -- > Guillaume Delhumeau ([email protected]) > Research & Development Engineer at XWiki SAS > Committer on the XWiki.org project > _______________________________________________ > devs mailing list > [email protected] > http://lists.xwiki.org/mailman/listinfo/devs > _______________________________________________ devs mailing list [email protected] http://lists.xwiki.org/mailman/listinfo/devs
