Thanks for the email Karl. Yes, when I read Laurie's concise and
improving the XHTML filters. I agree, these are bugs and we need to fix
them. I started the XHTML filters with 2 hopes: that those who use
filters would improve them. We all have 5 projects we're working on
easy and fun, is usually highest priority for each of us. I'm happy to
see a commit to improve the XHTML filters.
module.stripText(). This should give a reasonable plain-text
representation of a module entry.
Post by Karl KleinpasteLongish ramble.
I'm still finding our lack of attention or interest regarding
consistent output somewhat disappointing. David wrote a lot a couple
weeks back about this, but some of it just plain bugs me, and no one
else followed up at all. Some of the bugs-me is non-specific, some is
very specific.
Post by David "Judah's Shadow" BlueNow all this might sound pedantic and that front-ends should just
render what the engine sends, but imagine a frontend that sends the
text through a TTS engine for visually impaired persons. This
frontend would have no use for HTML formatting, but it would care
what the underlying markup that this HTML represents is.
I have 2 reactions to this observation.
1. I don't know if any other frontends are capable of it, but Xiphos
has been TTS-friendly for ~8 years. (Cf. Read Aloud in the View menu
for walking straight through a Bible, or using mouse-swept regions +
context menu invocation.) And indeed, as noted, I don't care what the
markup looks like, indeed I ship the text through a tag stripper
before it goes out to Festival. There is no consideration at all to
what was there, it is all blindly removed and simply shipped for
speaking. Yes, one could hypothetically say that a change of voice
could be used in the TTS case, but -- this is important -- _nobody cares_.
You see, for as long as I've been around, there has always been a huge
amount of talk within Sword about The Wonderful Things That Could Be
Done. But the real world's bottom line is that the five-9s proportion
of our actual user base has a straightforward goal: Good, productive
Bible study using quality tools. Precious few are deaf or blind, and
almost none are interested in full-tilt syntactic analysis tools. So
all the "see, the markup could let Joe Handicap have the text
delivered in his Special Way" really doesn't mean squat in the real world.
2. I'm not arguing against OSIS. Not all, in fact. Nor against
handicap support. But what I'll say about this sort of over-attention
to the hyperactive extremes of what could be done routinely leads us
to miss, or deliberately avoid, the underlying problem.
That problem, as addressed by Laurie Fooks' experimentation, is that
collectively we do a kind of poor job of consistent rendering of
what's under there. For the five-9s crowd, what they want is good,
consistent display of textual content. For the moment, ignore the
potential blind or deaf user of our apps. The problem is that we, the
frontend developers, have a pronounced tendency to produce /different
things/ for the /same text/, as displayed in the usual panes of our
apps. Is that because of the frontend itself, or because of what the
frontend gets from the engine?
That is where Laurie's experimentation reached, and where we
collectively fall over the cliff. We are accelerating at 32f/s^2
toward going */splat/* on the canyon floor below while blandly
discussing how nifty a more extensive TTS would be.
As far as the "not a Bible-reading program at all" crowd...I
absolutely do not care. The syntactic analyzer portion of our
userbase is past the nine-9s crowd.
It occurs to me that my "N-9s" nomenclature may be unclear or
unknown. Five-9s is 99.999%. Nine-9s is 99.9999999%. I use these in
mild hyperbole. When I say that the five-9s crowd wants just to see
good Bible textual display and whatever features can readily go with
that, I mean that out of a hundred thousand users, just one of them is
outside that usual space. Now, I said it's hyperbole, and someone
will tell me that they have a whole community of users for whom TTS is
crucial...totalling all of 10 people, or even 100 people. Well, fine,
you've moved the issue from five-9s to four-9s, 99.99%, and ...
whoopie. You have done nothing to make me feel better.
Besides, as noted, Xiphos has TTS, it's proven to be quite a bit more
welcome than I ever intended it to be -- it was my first significant
hack on then-GnomeSword, /which //I did //as an exercise /for code
familiarity's sake -- and to my knowledge there exists not one Xiphos
user who is a Xiphos user /because of /TTS. They are Xiphos users
because there are other aspects they like, having to do with workflow
or presentation or search capability or whatever else, and prefer
these aspects over other Sword apps, and TTS is a pleasantly useful
side capability they are happy to put to use.
...
The problem is that, for given OSIS constructs, the user -- and, let
us not be remiss, the publisher -- cannot be assured that what arrives
at the user's eyeballs is even readable, much less provided as intended.
Why is this? Xiphos uses straight engine output from XHTML filters.
BibleTime uses the engine but with its own filters. JSword apps have
their own entire separate engine, and JSword is now driving almost
half of all Sword apps out there. Within apps using the C++ engine,
BibleCS uses a different set of filters (RTF) which evidently do not
produce the same visual result as the XHTML filters. Consistency is
lacking because the end result of all these widely differing
filtrations cannot be counted upon. Is BibleCS' visual display
different because of inherent limitations or differences from Xiphos
vis-a-vis RTF vs. HTML display targets? I'm not in a position to
render an opinion on that.
But I can say that module authors like Laurie are stuck: They are
forced to use a least common denominator set of features of OSIS,
being the preferred markup methodology, because the interpretation of
that markup, as rendered into a display window, is not consistent,
cannot be counted upon.
Post by David "Judah's Shadow" BluePost by Laurie FooksIf we don't have a high level of commonality then I am concerned that
we are losing the purpose of having a common "engine"
Well, no not necessarily, the purpose of the engine is to read the
modules and then provide the content in a way that the front end can
meaningfully use for its purpose.
This makes me want to throw things.
The purpose of the engine is to produce the text according to the
filters which its client, the frontend, has chosen, and from which
that client expects consistency of output result.
That's all.
Anything else avoids the question. A frontend's "meaningful use" is
irrelevant if the engine's product is not in line with what the module
author wanted. Cf. Laurie's experience with tables and so forth. The
filters are supposed to produce output from an OSIS table spec,
however that looks in the original markup, into something that makes
sense in a frontend's HTML or RTF or WhateverElse widget. When those
filters don't do that, crud gets displayed, crud like Laurie found.
Post by David "Judah's Shadow" BlueYou can look at it this way, if the engine determined how frontends
should display the text, what would the point of multiple front ends be?
Wow, David, can you say "red herring"?
The reasons for picking a particular frontend have precious little to
do with that. What drives Xiphos users? Historically, for as long as
I've been involved with the code, it's been user interaction features
and visual quality, especially for user-created and -editable content,
which is why Xiphos has user annotations and editable genbooks and
tree-structured bookmarks and multi-bookmarks and verse highlights and
dynamically resizable images and a bunch of other things. That's what
users on the mailing lists have asked for, so that's what we've produced.
That the frontends should display the same text in closely the same
way should be axiomatic, and is precisely what Laurie is looking for.
And David whistled right past it. It's not textual display that
differentiates today's frontends, it's their other features far beyond
"can it display text properly?" We haven't even achieved that
level-zero axiom. We should be ashamed.
Besides this, I believe a hug/e, //HUGE/ aspect of this is bound up in
an issue that is nearly rejected outright, when it's given any
attention at all: How does The Publisher want it to look? Do we even
bother to ask? If we ask, do we listen to the answer? Laurie wants
OSIS tables to behave a certain way, namely, according to its own
spec. Has anybody taken her thoughts to heart? Have any filter fixes
been done in the name of accommodating what she perceives as bugs?
Post by David "Judah's Shadow" BlueBut until people need the additional features, they won't be added.
Images are a good example, this is very new to the code, but wasn't
put in until someone felt the need.
Images have been supported in the engine for as long as I've been
around, again about 8 years. This is not new by any reasonable
meaning of the word. By comparison, the XHTML filters are less than 2
years old and are already in wide use. There has been an ImageSampler
module since long before I was around.
Post by David "Judah's Shadow" BluePost by Laurie FooksI do feel that the system / project would benefit
from establishing (or publishing/clarifying same) minimum standards
for markup functionality
This is a laudable goal (with caveats for for the aforementioned
cases where presentation is not visual or doesn't care about
formatting for other reasons). But this is very difficult to do.
You might as well have said "we can't fix bugs."
The fact that Laurie can report that Feature X (such as table display)
doesn't work consistently in OSIS markup as rendered in all the
frontends means that the engines, both C++ and JSword, are buggy.
Mothing more and nothing less. They need to be fixed so that they do
what they're expected to do, so that they behave according to spec.
There /is/ a spec, y'know, a document that says there are minimum
standards for markup functionality, just as Laurie asked immediately
above. Once that's done right, if any further frontend bugs remain in
the resulting output as displayed, then we in frontend development
will have our own bugs to fix. Until then, it's an engine problem.
--karl
_______________________________________________
sword-devel mailing list: sword-devel at crosswire.org
http://www.crosswire.org/mailman/listinfo/sword-devel
Instructions to unsubscribe/change your settings at above page
An HTML attachment was scrubbed...
URL: <http://www.crosswire.org/pipermail/sword-devel/attachments/20141008/ee8d5b36/attachment-0001.html>