On Wed, Jun 10, 2015 at 6:36 PM, Ecaterina Moraru (Valica) <[email protected]> wrote:
Hi,
Since we discussed so many things lately and there are many technical constrain that we are facing, I'm going to try to get some conclusions from the UI perspective:
First of all there was the discussion about the usability tests. It's true that users are confused that we provide 2 levels of hierarchy (wiki/space/page and parent/child). If we can switch to just one it would be great since it will simplify the user's mental model. Another problem they had was with the naming, since wiki/space/page does not ring any folder/page bell, even with the icon usage. That's why I like the concept of nodes and only one type of entities that can have children. Since we are on the Web, I think the most likely entity name we could use is 'page', similar to content-page, web-page, wiki-page (even if technically is a space or a document).
Having only one type of entity, I agree to the fact that we should remove the space notion. We still have complexity, since we will have main-wiki, wikis*, pages* (child-pages).
I also agree to remove the notion of 'WebHome' from UI. We can still use it technically to differentiate old spaces from pages, but for the user doesn't mean anything, except looking technical.
Regarding the global-menu, we need to remove the wiki>space>page+actions display since it is not scalable. This means that the current breadcrumb zone will display the hierarchy. This is perfectly normal with the current navigation patterns used all over the Web. Note that classical breadcrumb do not display entity-type or actions. We shouldn't either. This means (as stated on http://design.xwiki.org/xwiki/bin/view/Proposal/ExtendedBreadcrumb) that we need to provide a Drawer + Move actions inside the content-menu.
Regarding actions, since we will have just one entity-type (not sure what we do with the wikis) we will need to standardize the entity actions. Currently we provide just 'Delete' for spaces, but if we don't display Spaces differently and they are just nodes, this means that we will need to implement or normalize the available actions. No matter what type it is technically, the user should be able to: Add (child), Edit, Delete, View, Administer, Copy, Rename, Move, Export, Watch, View History, View Informations, View Attachments, View Comments, View Rights, View Objects, etc.
Now, what I really don't like are the long URL. For me this is the most bad decision we can make. Nobody uses long URL on the web. I really like the way our current parent/child works, it provides the hierarchy in the UI, invisible in the URL (keeping it relevant to the current page). Hierarchy anyway should we displayed in Trees, not linear in URLs, since it's hard to read. People like short URL. They are easier to share, easier to scan, easier to understand what the content is about. We have the breadcrumbs for navigation, the URL is for identification.
+1, I don't like long URLs either. Thanks, Marius
Thanks, Caty
On Mon, Jun 8, 2015 at 11:40 AM, Eduard Moraru <[email protected]> wrote:
Hi,
Going on the "node" architecture, where we have only documents and child documents, here's something interesting that pops up as "the way to go" when storing a tree structure in SQL [1]: Closure Table [2]. It seems to cover pretty much all use cases, as solution 1 (Path Enumeration - what we are doing with the "space" document field) is limited [3] when it comes to changes in the tree structure (i.e. renaming/moving a document).
Note: There is also the extension [4] that includes a depth column which would be most useful as well. Basically, getting the list of parents of a document would cost only 1 query, sorting by depth to preserve hierarchy.
Of course, all this requires an extra table in the database for storing the hierarchy relationship.
WDYT?
Thanks, Eduard
---------- [1] http://www.slideshare.net/billkarwin/sql-antipatterns-strike-back/48 [2] http://www.slideshare.net/billkarwin/sql-antipatterns-strike-back/68 [3] http://www.slideshare.net/billkarwin/sql-antipatterns-strike-back/77 [4] http://www.slideshare.net/billkarwin/sql-antipatterns-strike-back/76
On Sat, Jun 6, 2015 at 3:38 PM, [email protected] <[email protected]> wrote:
On 6 Jun 2015 at 14:20:46, [email protected] ([email protected]
(mailto:
[email protected])) wrote:
Hi Marius,
On 4 Jun 2015 at 18:16:03, Marius Dumitru Florea ( [email protected](mailto:[email protected])) wrote:
On Wed, Jun 3, 2015 at 5:10 PM, Guillaume "Louis-Marie" Delhumeau wrote:
Hello XWiki committers.
Vincent have proposed the development of nested spaces for 7.2 and some of us have already agreed. But the concept of nested spaces introduces a problem that Denis have mentioned during some internal discussions at XWiki SAS, and that I am going to report here.
From a UI perspective, differentiating pages `A.B.C.WebHome` and `A.B.C` could become very difficult.
Moreover, we know that a lot of users do not understand the notion of spaces, and they are lost when you look at them during usability sessions.
I'm not sure about this statement. My feeling is that many users understand spaces as "folders" (they make the analogy with the file system). Moreover, whenever we display a space in the XWiki UI we use the folder icon, so we encourage the users to make this analogy.
I agree with Guillaume here. We’ve seen it on this list but more importantly Caty has done some recorded usability sessions and if I remember correctly it was clearly showing the problem. And users don’t understand that Spaces are like Folders which is why we also had this discussion on the list at one point about renaming them as Folder.
In any case, removing one concept can only be simpler IMO. I find it simpler to have 2 concepts (Wiki, Pages: A wiki is a set of pages) instead of 3 (Wiki, Spaces, Pages: a wiki is a set of spaces, which each one containing pages).
The situation is even worse if you consider the notion of parent/child documents, which is completely unrelated to the actual hierarchy. It creates confusion!
To fix these problems, we propose to introduce the notion of "nested documents", i.e. the ability to create documents inside documents.
What's the difference at a *conceptual* level between the notion of parent/child we have right now and the notion of nested documents you propose? I don't see it.
Yes it’s the same thing from a User POV. One difference is that we want a reference to contain the full path of a doc. Right now you’d need to transport the (Doc Reference + the full Breadcrumb) to represent the same thing.
But I agree that it’s the same concept, it’s just a different implementation. That’s why I’m proposing to drop the parent/child fields as we have them now since it’s just duplicating the concepts (and it’s confusing for users), and to reimplement the Breadcrumb UI using Doc References.
Note that having the full location in the reference also allows to have URLs containing the full path which is interesting (for knowing where you are by looking at the URL: http://.../view/Space.Page1, http:// .../view/Space.Page2 doesn’t indicate anything about the relationship between Page1 and Page2 when Page1 could the parent of Page2). It’s also interesting for SEO.
You even use "parent" and "child" words below to explain the "nested" notion. The word "nested" sounds very technical to me. I don't know about French, but in my native language the translation for "nested" is not a commonly used word, unlike parent and child. It seems easier to me to explan to a user that a document can have a parent document and some child documents (the parent / child relationship) then to explain them that a document can be "nested" inside another document and can "nest" other documents too.
I don’t think GL has suggested to use the word Nested anywhere in the UI… It’s just a word we use internally to describe the feature. I’m fine to continue using the terminology Parent and Children.
Say differently, if a page `A.B.C` exists, nothing should stop the user to create the document `A.B.C.D`.
You mentioned JCR on the next paragraph. Are JCR nodes identified by the position (path) in the tree?
Yes. They call it Path (we call it a Document Reference).
I think we should make a distinction between the way we identify a document and the way we access that document.
This means using unique ids for identifying docs (and this is what we already have in the DB except the Id is based on its path in the DB but this could be changed and this is our problem). Of course in the UI we should never display these ids.
I like the fact that currently when you change the parent of a document the document identifier (reference) stays the same.
It’s not fully true. When we rename a doc the doc id is modified.
What would be interesting would be the ability to have several Document References for a given doc (with one being the default probably). This is something I’ve been interested in implementing for a long time but it would probably require some model changes and it’s not for XWiki 7.x IMO. We could discuss it when we talk about XWiki 8.x and the new model in general. See also http://design.xwiki.org/xwiki/bin/view/Design/XWikiModel20 (on that page I had put: “Ability to have multiple references pointing to the same entity” and yes this is also supported by the JCR API).
Also note the idea of "An Entity can be a pointer to another Entity (e.g of use case: rename, aliases)” from that doc. This should be not to hard to implement using our current model by adding a Type field in the DB. But it’s a breaking change that would require 8.x (because current apps doing queries on the doc table would get the “pointer” Types which should be excluded, unless we add a filter to search*() APIs + the Query API).
Thanks -Vincent
In JCR[1], there is only one concept: the "node". A node can have a content, and a list of child nodes. In XWiki, documents could become a kind of nodes, and we do not need spaces anymore.
If we don't have space anymore, we could ask ourselves: "How the rights will be propagated to the children documents? How do we distinguish rights applied to the documents and the rights applied to the children?"
I think the easiest solution is to inherit the rights from the parent to the children, unless an object prevent it. We already have this kind of mechanism with XWikiRights and XWikiGlobalRights. XWikiRights would be applied for the current document, and XWikiGlobalRights for the document and its children.
But changing the XWiki model is a lot of work, that we don't have time to achieve for 7.2. So we propose to make it step by step.
The first step is to change the UI to hide the notion of space to the users. Concretely, each time a user wants to create a page called `A`, we actually create the document `A.WebHome`. So any child of this page would be created in the `A` space, like `A.Child`. But this child would be in a space too, so it would be `A.Child.WebHome` actually.
I find this 'A.WebHome' thing too complex. Look at the document hierarchy tree
http://extensions.xwiki.org/xwiki/bin/view/Extension/Document+Tree+Macro#HDo...
. We can hide the spaces already by relying on the parent / child relationship.
A.WebHome is too complex which is exactly why Guillaume is sending this proposal about Nested Documents. The idea is that the doc will be seen as “A" for the user (and implemented technically as A.WebHome till we update the DB to remove the “space” field, which would make the impl much simpler).
Then, when we display the `A.WebHome` page, we remove all mentions to the `WebHome` name. In the UI, it will just be presented as the document `A`. This is a good point, knowing the fact that the term `WebHome` have no sense for the user, neither in English or in other languages.
Again, these changes are only for the UI. For the applications, it is the developer's job to decide if the app will create documents like `Document.WebHome` or basic documents just as before.
The question of what to do with AWM comes up. When a user creates an entry, should it be a new-kind-of-document (`AppSpace.Entry.WebHome`) or an old-kind-of-document (`AppSpace.Entry`)? The first option is good for consistency and for the new possibilities it offers, but the second is better for retro-compatibility. And the question will be the same for all existing applications that create pages. I believe we should answer these questions on a case-by-case basis and deserve their own mail threads.
This proposal also implies to change some macros, like the {{space}} one, and some panels. But I believe there is no blocking-point there.
Finally, after these steps are accomplished in 7.2 and polished until the end of the 7.x cycle, we will refactor the XWiki model (something we dream about for years).
To sum up, the idea we propose is:
On the short run:
- Hide the notion of space in the UI.
- Hide the `WebHome` name in the UI.
- When a user creates a page from the UI, it actually creates a space with a WebHome.
- Remove the current parent/child mechanism which is outdated (and confusing) compared to the new hierarchy.
On the long run:
- Remove the notion of space in the model, and replace it by "nested documents".
- Tune the rights system to inherit rights from parents to children.
Of course, we can discuss the technical details and the implementation strategies. But for now, we need to know if you accept the general idea (nested documents).
So, I hope you will like this proposal, and here is my +1.
I'm not convinced. I don't see why we should drop the parent/child relationship in order to introduce something similar but more complex.
I’d say mostly because we need to transport the full location in the Doc Reference, which allows us to do lots of things: - display it in the URL - no need to recompute the breadcrumb every time we need to display the hierarchy (which is the case now and it doesn’t scale) - ability in the future to have several references for a single doc - consistency: there’s no reason that the space would be in the reference but not the parent… ATM we have two concurrent concepts which makes it impossible for users to understand the difference between Spaces and child/parent relationships.
Thanks -Vincent
Thanks, Marius
Thanks,
Guillaume D.
[1] JCR:
https://en.wikipedia.org/wiki/Content_repository_API_for_Java
-- Guillaume Delhumeau ([email protected])
_______________________________________________ devs mailing list [email protected] http://lists.xwiki.org/mailman/listinfo/devs
_______________________________________________ devs mailing list [email protected] http://lists.xwiki.org/mailman/listinfo/devs
_______________________________________________ devs mailing list [email protected] http://lists.xwiki.org/mailman/listinfo/devs