See also: IRC log
rssagent, pointer?
scribenick bjdmeest
https://github.com/w3c/dpub-pwp-loc/issues/9
ivan: about arbitrary resource within a
packed PWP
... if no unpacked version, this will give a 404
lrosenth: given the server cannot process the packed PWP
ivan: this proposal is algorithmically ok
... but not realistic to set up
lrosenth: or the server could dynamically return the resource
ivan: then you wouldn't get a 404
lrosent: so you need a PWP-aware server
ivan: could be done via .htaccess
lrosenth: it is possible
ivan: viable, but not realistic
bjdmeest: so nothing more than a 404?
ivan: we could be a little more explicit, but it's more theoretical
https://github.com/w3c/dpub-pwp-loc/issues/10
ivan: currently, the document says the complete PWP consists L and Lu and Lp
lrosenth: we weren't going to talk about the content of the M
ivan: we don't say anything more than about
the locators
... state-dependent locators are e.g., Lu and Lp, locators to unpacked
and packed version
lrosenth: how can that remain state during
the lifetime of PWP?
... when I create a PWP, I assume my PWP has a manifest
... once built, it cannot change
... there may be extra Ms
... but you cannot change the initial M
... the locator in the initial M is the locator of the original source
ivan: when the PWP processor gets the full
PWP manifest (maybe from different places)
... in that manifest, there are the different locators, and the
canonical locator
lrosenth: so definition is about final manifest
ivan: yes
... so it is already in the draft
bjdmeest: great, let's close 10
http://w3c.github.io/dpub-pwp-loc/#algorithm-to-find-the-right-values-for-lp-and-lu
https://github.com/w3c/dpub-pwp-loc/pull/24
about issue #21
ivan: the original intro of the algorithm
stated that it gets the manifest, as a consequence, it retrieves L*
... I re-edited that part
<ivan> https://github.com/w3c/dpub-pwp-loc/pull/24/files
rdeltour: this is about finding the
manifest, if final spec says that L == Lu, then we already know Lu, so
part of the algorithm would be pruned?
... we don't say nor exclude
... I'm not sure it's a common use case to know both Lp and Lu
lrosenth: it's not
rdeltour: so why talk about it?
lrosenth: you need to get one of them
rdeltour: if it's an OR, given that in some cases you get Lu because L == Lu, then you don't need this algorithm
ivan: the PWP cannot make this assumption
... the final spec might be very restricted
... that changes things, but I would be very much against such a
restriction
... so currently, the PWP processor cannot assume such a restriction
rdeltour: given we already need the manifest, then let's talk about this section as 'getting the manifest'
ivan: in a general doc, you are probably right, but right now we are talking about the locators
rdeltour: but the manifest might have other information about the relative locators, so this is not just about Lu and Lp
ivan: this is highly editorial, at some
point there is a more general discussion about what to do with this
document
... in such a general document, the section of dpub-loc would probably
be moved to the general document
... atm, we are not at that level yet
... that's why I made the goal of this section: finding L/Lu/Lp via
finding the manifest
rdeltour: I think as soon we talk about
internal resources, then we need to get back at the manifest, and need
that section as well
... so why hide that this section is about fetching the manifest?
lrosenth: so let's change the title
ivan: fine, I will move this section as a
top-level section
... later, we can see how this evolves
... so this is about: what kind of document will we end up with? this is
not just locators
... but happy to make the change
https://github.com/w3c/dpub-pwp-loc/pull/25/files
ivan: added a note about remaining silent about retrieving a manifest from a package
rdeltour: that's ok, but comment about
package being a black box... in the note, for some part of it, we go
into very technical details (e.g., for html)
... but we don't for package.. it makes the whole thing
detailed/non-detailed
... maybe large parts of the algorithm may be pruned
... I see the algorithm more as an example of a complex situation
... I think the note should be short and make general statements
ivan: this is similar as before: we don't
know where this document will go
... the only thing we do, is collect inputs, that would be input into a
WG, that standardizes PWP
... and all documents would need to be properly redone
... atm, we have a good idea, so documenting as precise as we can,
that's helpful for future WG
... e.g., terminology discussion, that can change by a WG, but we
documented that
... we came up with this rough idea about how things could be done
... and we wrote it down properly
rdeltour: I think it's great to have it
... but I have a problem with the complexity on the one hand and the
vagueness on the other hand
... I would feel a short list of general statements would help
... also, the entire section sounds like we nailed it, and have a formal
spec
... but there are still unresolved things
... so: 1) perceived complexity, 2) perceived finality
bjdmeest: so we need a better intro? adding those short statements, that this is not final, etc.
ivan: that will be done because it will be
separate section
... I would merge the pull requests, and another round of editing, and
then look at the results
... [about the complexity] eventually, a bunch of details will be
described in the same level of complexity
lrosenth: in order to accomplish all of the goals, it is the reality
rdeltour: so final specification could only provide some alternatives
ivan: if you look at the EPUB discussion..
rdeltour: depending on the eventual spec, the algorithm might be simpler
lrosenth: or more complex
... all depends on the eventual implementation
https://github.com/w3c/dpub-pwp-loc/pull/26
rssagent, pointer?
<rdeltour> +1
Topic #6
<scribe> closed
Topic #11
https://github.com/w3c/dpub-pwp-loc/issues/11
lrosenth: same as what happens when you download a PWP
rdeltour: when you have a PWP online and
download it, you have the canonical, dereferencable locator
... if I have created a PWP locally, I don't have a canonical locator
lrosenth: you might have, because if the
PWP must include L, every created PWP will have a L, but L might not be
dereferencable
... there is no requirement that the uri of a XML namespace must be
dereferenced, maybe the same could go for locators of PWP
rdeltour: but we always talked about URLs, not URIs
lrosenth: if we only have a
email-distributed PWP, and only read locally
... it should still work
ivan: I would still like to have a clearer use case
<rdeltour> +1 on the need for a UC
ivan: an HTML send around by e-mail... that HTML file is not on the web
lrosenth: do PWP has to be on the web?
ivan: I don't know
lrosenth: UC: before I'm ready to publish
it
... during the authoring process, I create drafts, and send PWPs via
emails to reviewers
... so it doesn't exist on the web, but is built as a PWP
ivan: so your reviewer will receive you book, on local filesystem (file:///home/book.pwp), and thus has a locator
lrosenth: so different locators
... so not necessarily in the file, but in the combined manifest
ivan: we need more concrete use caess
rdeltour: kind of included in portability buckets
ivan: next week, i might have to skip
rssagent, draft minutes
<ivan> Title: DPUB IG Locator TF call