[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: Re: [dita] Potential Issue: When Does Applicability Apply to KeySpace Determination?
I think that reasoning makes sense in general and it also makes sense for the Toolkit that the natural implementation decision would be to put any new processing, including keyref, after filtering. That seems to argue the general principle that filtering is first. I don't know that that principle is formally stated or clearly implied anywhere in the current spec. So I guess we really have two questions: 1. Does anyone disagree with the assertion that "filter first" is/should be the working principle of the DITA 1.2 standard? 2. If everyone agrees that "filter first" is the principle to apply to processing generally, does anyone disagree with the conclusion that key spaces must therefore reflect the effect of filtering? If the answer is "no" (no disagreement), then the action would be to update the language in "Indirect content inclusion: keys and conkeyref" under the "Linking and addressing" topic to include an explicit statement that filtering is applied before determining the key space and that the effective key space may include multiple definitions of the same key with different applicabilities. To recap, the implementation implication for "filter first" on systems that need to be able to calculate key spaces that reflect all possible sets of conditions used in a given information set is: 1. Either take a condition set (e.g., DIVAVAL document) as a parameter to the keyspace construction or 2. Return a keyspace that reflects not just unique keys but unique key/condition set pairs, meaning that there may be multiple instances of the same key reflecting different, unique sets of conditions. Note that by "keyspace construction" is implicitly meant also "key-to-resource binding lookup" since you can't resolve a key to a resource until you have first constructed the keyspace. The key mechanism already requires a root map as a parameter to the keyspace construction process. And again, this is primarily a practical issue for authoring and retrieval, not rendition processing. In general I think of authoring as involving both an authoring tool (editor) and a "link management service" that supports the editor by optimizing knowledge of links and addresses, in particular, what the set of available targets are (think I can address and therefore link to) and, for a given thing that *could* be addressed, that things do in fact address it (where used). I call this a "link management service", rather than content management, because it could be implemented as a standalone service applied to data that is otherwise not managed (e.g., topics on a file system or in a code control system or whatever) or it could be built into the editor (e.g., OxygenXML's built-in support for keys). However, one would normally expect link management services to be provided by more expansive content management systems that also provide versioning, access control, searching, and so on. Cheers, Eliot On 11/2/09 3:40 PM, "Robert D Anderson" <robander@us.ibm.com> wrote: > Hi Eliot, > >> I see that the current 1.5 Toolkit does filtering before doing any other >> processing. Is that a considered decision or just the way the > implementation >> fell? > > It was a considered decisions with regards to all other aspects of DITA > processing, though we did not re-examine that decision in light of keyref. > The reasoning has been discussed on this list and others, but it is > primarily based on experiences that the IBM team had in the early > (pre-standard) days of DITA. The two primary reasons to do it this way are > 1) improve processing speed by dropping content you won't use as soon as > possible, and more importantly 2) do not try and retrieve content that will > never be needed. For example, if I have this: > <p conref="productA-info.dita#topic/p" product="a"/> > > If I want to filter out all product A information, then I drop this early > on in my processing. The build does not retrieve productA-info.dita, which > may not even be available. The conref processing is also skipped, which > saves a bit of time. > > The bigger issue for many of my own users is that when they package up > information to send off to a translator, they do not need to ship > productA-info.dita, because the conref is filtered out and never resolved. > So, no money is wasted translating the file that will not be used. This > issue, rather than the slight processing speed improvement, was the main > one that resulted in the current processing order. > > As I said, when keyref came around, we didn't step back to re-address that > issue, but it does make me wonder - if you have a key definition that wins > with no ditaval, but is filtered out when you try to render the content, > how could that key still be valid in your rendered files? I guess that's > partly what you're asking. At this point I would expect the results you're > getting from the OT today, but I'm obviously thinking about this from an OT > style publishing perspective rather than from an editor or CMS perspective. > > Robert D Anderson > IBM Authoring Tools Development > Chief Architect, DITA Open Toolkit > > ekimber <ekimber@reallysi.com> wrote on 11/02/2009 02:25:28 PM: > >> ekimber <ekimber@reallysi.com> >> 11/02/2009 02:25 PM >> >> To >> >> dita <dita@lists.oasis-open.org> >> >> cc >> >> Subject >> >> [dita] Potential Issue: When Does Applicability Apply to Key Space >> Determination? >> >> Something that came up in the context of a discussion I was having on how >> one would normally process DITA content in the context of keys led to the >> question of when conditional processing is applied to key definitions. >> >> In particular, is filtering applied before or after the key space is >> determined for a given map? >> >> If filtering is applied before the keyspace is determined, it means that > you >> cannot determine a key space given just a map: you must also specify a >> DITAVAL file (or its equivalent). It also means that the same map tree > may >> produce different key spaces for different DITAVAL files. Or rather, it >> means that a key space is not simply a unique set of keys, but a unique > set >> of key/property pairs (where the same key may occur more than once as > long >> as each instance has distinct applicability). >> >> If filtering is applied after the keyspace is determined, then you cannot >> have key definitions that use conditional processing. >> >> From an implementation standpoint, applying filtering *before* keyspace >> determination significantly complicates key space calculation for systems >> that just deal with keys (e.g., authoring systems) and are not doing >> sequential processing of DITA content. >> >> If filtering is done before keyspace determination it means that all >> processors that work with key spaces must take as a parameter to any key >> lookup both the root map *and* the set of applicable conditions or else > they >> have to include in any set of available keys all key definitions that > have >> unique applicability, that is, for a given key, the first of each > definition >> with a unique set of @props values, that is, the set of definitions that >> *could* be effective. This would definitely complicate key space >> calculation, but it could be done if the requirement is understood up > front >> and systems are designed to accommodate it. >> >> If filtering is done after keyspace determination, it means that > conditions >> don't affect key definition (that is, a key-defining topicref that would > be >> filtered out is not filtered out for the purpose of defining keys). >> >> I see that the current 1.5 Toolkit does filtering before doing any other >> processing. Is that a considered decision or just the way the > implementation >> fell? >> >> From a CMS and authoring tool implementation standpoint doing filtering >> *after* key space determination would be easiest to implement but I > suspect >> that that is too limiting as it denies use of filtering for key > definition >> management. >> >> Do we already have a definitive answer to this question? If not, is there >> consensus about what the right answer is? >> >> Thanks, >> >> Eliot >> >> ---- >> Eliot Kimber | Senior Solutions Architect | Really Strategies, Inc. >> email: ekimber@reallysi.com <mailto:ekimber@reallysi.com> >> office: 610.631.6770 | cell: 512.554.9368 >> 2570 Boulevard of the Generals | Suite 213 | Audubon, PA 19403 >> www.reallysi.com <http://www.reallysi.com> | http://blog.reallysi.com >> <http://blog.reallysi.com> | www.rsuitecms.com <http://www.rsuitecms.com> > >> >> >> --------------------------------------------------------------------- >> To unsubscribe from this mail list, you must leave the OASIS TC that >> generates this mail. Follow this link to all your TCs in OASIS at: >> https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php >> > ---- Eliot Kimber | Senior Solutions Architect | Really Strategies, Inc. email: ekimber@reallysi.com <mailto:ekimber@reallysi.com> office: 610.631.6770 | cell: 512.554.9368 2570 Boulevard of the Generals | Suite 213 | Audubon, PA 19403 www.reallysi.com <http://www.reallysi.com> | http://blog.reallysi.com <http://blog.reallysi.com> | www.rsuitecms.com <http://www.rsuitecms.com>
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]