Re: [sword-devel] Catholic versification / inter-versification mappings
ycliffe use - even when allowing for verse bridges, reordering and more. Which makes me think we should think more about a presentation overlay which we have at the moment very rudimentary only. OSIS certainly allows for more than we incorporate right now. This could take eg in the engine the form of a resorting filter where text is reordered and marked up for user facing presentation based on additional presentation info in modules. This would to my mind be a much better solution than ever increasing further system of av11n or indeed a complete rewrite. Peter Sent from Outlook for iOS <https://aka.ms/o0ukef> -------------------- *From:* sword-devel on behalf of Troy A. Griffitts *Sent:* Tuesday, February 20, 2024 12:24 am *To:* sword-devel@crosswire.org *Subject:* Re: [sword-devel] Catholic versification / inter-versification mappings Dear all, These comments are a mix of background, history, and thoughts: 1) VERSIFICATION (v11n): Variation between reference systems sucks. Until you get into the weeds of the details, it is normal to assume the problems are not complex. SWORD tries to implement a simple 90% solution. SWORD and JSword support defined abstract versification schemes with 3 simple dimensions: [bookid : chapterMax][chapterNumber : verseMax][verseNumber : verseEntry] Conceptually we also operate on these assumptions (I've skimmed the proposal by Arnaud which differs here, but I haven't given it the thought it deserves to comment yet): that book order is defined in the v11n system; that chapter and verse numbers are numeric and begin at 1 and increase to verseMax. We also allocate a special slot '0' for: Module Introduction; Testament Introduction; Book Introduction; and Chapter Introduction (e.g., Matt.0.0 can hold an introduction to Matthew). Those who have been exposed to many Bibles will immediately think of places these assumption fall short. But for >90% of our Bibles, these assumption hold true, and these assumption make many aspects of our work much simpler (abstract parsing of verse lists and ranges, bookmark ordering, etc.). Historically, SWORD previously supported dynamic, per module, versification, with a 3 phase lookup: index file .bks[book number] = book offset in next index; index file .cps[book offset + chapter number] = chapter offset in next index; index file .vss[chapter offset + verse number] = verse offset and entry size in data file. 20 years or so, we made the decision to begin the hard work to understand versification systems within Bibles so we could begin to map them appropriately. This let us remove the .bks, and .cps index files and store that data in versification system definitions, leaving only the final .vss index file which gave the offsets and entry sizes into the data file. Caring about versifications was a decision we made. Our previous design let any Bible decide how many books, how many chapters, and how many verses each chapter contained. This had its merits because any new versification could be defined in each module without anyone caring what it was. But the drawback was the same: any Bible could decide how many books, how many chapters, and how many verses without anyone knowing why or what they were. Some have pushed for dynamic definitions of v11n systems again, and I understand why. I am in favor of moving forward with a hybrid approach: a set of defined versification systems, which a module will still need to choose from, to which it most closely adheres, + the ability for that module to specify its variation. Toward 98%: We have tried to work around the cons of this simple design and approach 100% support by accounting for the most common types of problems, e.g. * The engine allows common verse suffixes (e.g. Matt.2.7b); * The engine skips verses in a Bible which are not present-- this allows us to create v11n schemes which are a superset of n number of closely related v11n schemes, knowing that the engine will skip over the verses that are not present in the module; We also have tools which print out missing verses which has proven a good QA check for our modules team. * When we run across a Bible which adds an odd verse here or there or an out of order verse, our workaround has been to append these to end of the verse just before where they should appear, so the text
Re: [sword-devel] Catholic versification / inter-versification mappings
e eg in the engine the form of a resorting filter where text > is reordered and marked up for user facing presentation based on additional > presentation info in modules. > > This would to my mind be a much better solution than ever increasing > further system of av11n or indeed a complete rewrite. > > Peter > > > Sent from Outlook for iOS <https://aka.ms/o0ukef> > > -- > *From:* sword-devel on behalf of Troy > A. Griffitts > *Sent:* Tuesday, February 20, 2024 12:24 am > *To:* sword-devel@crosswire.org > *Subject:* Re: [sword-devel] Catholic versification / inter-versification > mappings > > > Dear all, > > These comments are a mix of background, history, and thoughts: > > 1) VERSIFICATION (v11n): > > Variation between reference systems sucks. Until you get into the weeds > of the details, it is normal to assume the problems are not complex. SWORD > tries to implement a simple 90% solution. > > SWORD and JSword support defined abstract versification schemes with 3 > simple dimensions: [bookid : chapterMax][chapterNumber : > verseMax][verseNumber : verseEntry] > > Conceptually we also operate on these assumptions (I've skimmed the > proposal by Arnaud which differs here, but I haven't given it the thought > it deserves to comment yet): that book order is defined in the v11n system; > that chapter and verse numbers are numeric and begin at 1 and increase to > verseMax. We also allocate a special slot '0' for: Module Introduction; > Testament Introduction; Book Introduction; and Chapter Introduction (e.g., > Matt.0.0 can hold an introduction to Matthew). > > Those who have been exposed to many Bibles will immediately think of > places these assumption fall short. But for >90% of our Bibles, these > assumption hold true, and these assumption make many aspects of our work > much simpler (abstract parsing of verse lists and ranges, bookmark > ordering, etc.). > > Historically, SWORD previously supported dynamic, per module, > versification, with a 3 phase lookup: > index file .bks[book number] = book offset in next index; > index file .cps[book offset + chapter number] = chapter offset in next > index; > index file .vss[chapter offset + verse number] = verse offset and entry > size in data file. > > 20 years or so, we made the decision to begin the hard work to understand > versification systems within Bibles so we could begin to map them > appropriately. This let us remove the .bks, and .cps index files and store > that data in versification system definitions, leaving only the final .vss > index file which gave the offsets and entry sizes into the data file. > > Caring about versifications was a decision we made. Our previous design > let any Bible decide how many books, how many chapters, and how many verses > each chapter contained. This had its merits because any new versification > could be defined in each module without anyone caring what it was. But the > drawback was the same: any Bible could decide how many books, how many > chapters, and how many verses without anyone knowing why or what they were. > > Some have pushed for dynamic definitions of v11n systems again, and I > understand why. I am in favor of moving forward with a hybrid approach: a > set of defined versification systems, which a module will still need to > choose from, to which it most closely adheres, + the ability for that > module to specify its variation. > > Toward 98%: We have tried to work around the cons of this simple design > and approach 100% support by accounting for the most common types of > problems, e.g. > >- The engine allows common verse suffixes (e.g. Matt.2.7b); >- The engine skips verses in a Bible which are not present-- this >allows us to create v11n schemes which are a superset of n number of >closely related v11n schemes, knowing that the engine will skip over the >verses that are not present in the module; We also have tools which print >out missing verses which has proven a good QA check for our modules team. >- When we run across a Bible which adds an odd verse here or there or >an out of order verse, our workaround has been to append these to end of >the verse just before where they should appear, so the text flows the same >as the printed Bible, and we include for the reader an inline visual >separator and marker showing the publisher's verse number. > > > These work arounds get us pretty close to being able to support 98% of our > Bibles exactly as the publisher wishes, and the remaining 2% is supported > "well enough" for no complaints by publishers. Could we build a system > which allowed out of order verses, or which allow
Re: [sword-devel] Catholic versification / inter-versification mappings
There are two aspects here:Historical developments which have been by and large well explored by Chris and are mirrored well in our various av11n system - with specific and now largely known gaps - we have less clue on the Roman catholic development of versification systems and probably even less on a variety of autochthone churches with their own historical development of translations an versification. But these are well defined tasks and e.g. the work of Dominique to bring in the French systems documents well how we should handle that side. The complications coming in due to modern translation are actually very limited:FWIW, most modern translations which play about with versifications still actually follow an underlying well known and documented plan. NRSV(A) eg seems to be all what UBS and Wycliffe use - even when allowing for verse bridges, reordering and more. Which makes me think we should think more about a presentation overlay which we have at the moment very rudimentary only. OSIS certainly allows for more than we incorporate right now. This could take eg in the engine the form of a resorting filter where text is reordered and marked up for user facing presentation based on additional presentation info in modules. This would to my mind be a much better solution than ever increasing further system of av11n or indeed a complete rewrite. Peter Sent from Outlook for iOS From: sword-devel on behalf of Troy A. Griffitts Sent: Tuesday, February 20, 2024 12:24 amTo: sword-devel@crosswire.org Subject: Re: [sword-devel] Catholic versification / inter-versification mappings Dear all, These comments are a mix of background, history, and thoughts: 1) VERSIFICATION (v11n): Variation between reference systems sucks. Until you get into the weeds of the details, it is normal to assume the problems are not complex. SWORD tries to implement a simple 90% solution. SWORD and JSword support defined abstract versification schemes with 3 simple dimensions: [bookid : chapterMax][chapterNumber : verseMax][verseNumber : verseEntry] Conceptually we also operate on these assumptions (I've skimmed the proposal by Arnaud which differs here, but I haven't given it the thought it deserves to comment yet): that book order is defined in the v11n system; that chapter and verse numbers are numeric and begin at 1 and increase to verseMax. We also allocate a special slot '0' for: Module Introduction; Testament Introduction; Book Introduction; and Chapter Introduction (e.g., Matt.0.0 can hold an introduction to Matthew). Those who have been exposed to many Bibles will immediately think of places these assumption fall short. But for >90% of our Bibles, these assumption hold true, and these assumption make many aspects of our work much simpler (abstract parsing of verse lists and ranges, bookmark ordering, etc.). Historically, SWORD previously supported dynamic, per module, versification, with a 3 phase lookup: index file .bks[book number] = book offset in next index; index file .cps[book offset + chapter number] = chapter offset in next index; index file .vss[chapter offset + verse number] = verse offset and entry size in data file. 20 years or so, we made the decision to begin the hard work to understand versification systems within Bibles so we could begin to map them appropriately. This let us remove the .bks, and .cps index files and store that data in versification system definitions, leaving only the final .vss index file which gave the offsets and entry sizes into the data file. Caring about versifications was a decision we made. Our previous design let any Bible decide how many books, how many chapters, and how many verses each chapter contained. This had its merits because any new versification could be defined in each module without anyone caring what it was. But the drawback was the same: any Bible could decide how many books, how many chapters, and how many verses without anyone knowing why or what they were. Some have pushed for dynamic definitions of v11n systems again, and I understand why. I am in favor of moving forward with a hybrid approach: a set of defined versification systems, which a module will still need to choose from, to which it most closely adheres, + the ability for that module to specify its variation. Toward 98%: We have tried to work around the cons of this simple design and approach 100% support by accounting for the most common types of problems, e.g. The engine allows common verse suffixes (e.g. Matt.2.7b); The engine skips verses in a Bible which are not present-- this allows us to create v11n schemes which are a superset o
Re: [sword-devel] Catholic versification / inter-versification mappings
Michael and I obviously differ in opinion on this subject, and he knows that I love him and so appreciate the work he does at ebible.org. I would like to propose that our differences are due to our perspective in light of our primary ministries: Among many other things, Michael with ebible.org edits and publishes the World English Bible and its variations and also transforms every freely distributable Bible from the United Bible Societies' (USB) Digital Bible Library (DBL) repository to many formats, including SWORD, and for that we are so grateful and blessed by his efforts. For his work, he needs to try to fit all these Bibles into one of our versification schemes (actually maybe he has punted on that and may use our NRSVA scheme which is one of our largest supersets; I don't know these days). I can imagine the frustration this must be for him, trying to automate this conversion as much as possible. He has had much grace to convert DBL markup to OSIS (a format he doesn't love) and select a versification, which he sees many of his Bibles don't match exactly. I am so grateful for his friendship and his partnership in ministry over many decades. From my perspective, SWORD is focused on Bible research. In my mind, I am need to define a finite set of unambiguous segments of Scripture, and once defined, for each segment: * show that same segment of Scripture in all Bibles which contain that segment. https://crosswire.org/study/parallelstudy.jsp?del=all=KJV=NA28DBG=NASB=RST=Matt.16.28#cv * even try to align individual words, if we can, within that segment (try clicking on any word from the link above). * find the precise folio in all ancient manuscripts of that segment and showing images of that segment: https://crosswire.org/study/fetchdata.jsp?mod=TC=Matt.16.28=NASB * find the parallel passages of that segment across the Four Gospels, using Eusebian Canon Tables, with 2 segments of context before and after: https://crosswire.org/study/eusebian.jsp?key=Matt.16.28=2#cv * find and show commentary about that segment: https://crosswire.org/study/passagestudy.jsp?key=Matt.16.28-28=JFB * compute and show variation in manuscripts about that segment: * https://ntvmr.uni-muenster.de/community/vmr/api/collate/?baseText=ECM=Matt+16%3A28=0=-1=true=true=true=true=graph=undefined=intfadmin For my daily work, I need tightly defined, unambiguous segments of the original Text. Michael is daily trying to get a publisher's translation displaying the way the publisher desires-- with all the publisher's unique adjustments of the text, and understandably he gets frustrated that a simple task isn't so simple in SWORD because he needs to segment and fit each Bible into a versification system with IDs which we can use to do all our research. I get it. I understand. Simple things should be simple, and right now they are not. We have discussed alternatives for a more ePub/PDF like reader's view of a Bible, where we preserve the imported data file's ordering such that a reader oriented view could possibly ask for, say 'chapter' chunks and just display what it gets, and then slice things up into our indices behind the scenes as we do now. We've had 'reader view' requests for our lexicon modules, as well, because we order entries to allow fast binary search lookups, and when a client of the SWORD library asks to step through a lexicon module, they get an ordered list of entries instead of a the published order in the original lexicon. We could do much better preserving original publisher display; even simply to the point that we support this use case as a first class citizen in the engine and preserve a ePub or / PDF of the original with each module. I want to move forward with solutions for each task we each individually feel called to work toward. I am sorry that I prioritize my tasks I feel I am called to work toward instead of helping others find solutions for their ministry. This might be normal for a software engineer, but not for a good leader. Serving together, Troy On February 19, 2024 18:12:55 MST, Kahunapule Michael Johnson wrote: Dear All, If I understand Arnaud correctly, I really like his ideas. The BEST part is that the next time a Bible is submitted for processing with yet another unique versification (after the changes are implemented), it doesn't have to be either force-fit into a versification that doesn't fit or wait for decades for someone to update the hard-coded versifications in the Sword engine, and for those to be incorporated into all of the front ends. I regard the current minimalist versification system to be seriously in need of an upgrade. It is based on false assumptions (listed by Troy, no offense intended) that seemed good at the time they were made. However, with 1404 Bible translations (and counting) is that (1) 90% success is an over-estimate of how well
Re: [sword-devel] Catholic versification / inter-versification mappings
Dear All, If I understand Arnaud correctly, I really like his ideas. The BEST part is that the next time a Bible is submitted for processing with yet another unique versification (after the changes are implemented), it doesn't have to be either force-fit into a versification that doesn't fit or wait for decades for someone to update the hard-coded versifications in the Sword engine, and for those to be incorporated into all of the front ends. I regard the current minimalist versification system to be seriously in need of an upgrade. It is based on false assumptions (listed by Troy, no offense intended) that seemed good at the time they were made. However, with 1404 Bible translations (and counting) is that (1) 90% success is an over-estimate of how well it works, and (2) Sword versification is a complete failure for numerous projects because none of the existing versifications fit, the fall-back mechanisms fail and result in wrong outputs or crashes in osis2mod, and nobody is actively fixing the situation. I have found the following to be true: The number of versifications needed to represent all Bibles properly tomorrow is highly likely to be more than the number that works today. Hard-coding versifications into slowly-changing code that is only updated in fits and starts is doomed to fail (and already has, in my not-so-humble opinion). Verse numbers in a chapter don't always proceed in numerical order. Several Bible translations move the statement about the motion of the shadow on Hezekaiah's steps to a more logical place in terms of discourse, without changing the verse numbers. Indeed, they split verses into segments and straddle other verses with them. Chapter and verse "numbers" aren't always pure numbers. Letters get involved in the Deuterocanon/Apocrypha. Some Bible translators like to use verse segments (like 6a and 6b) heavily. Verse bridges (like verse 1-3 with everything from verses 1 through 3 but possibly rearranged and with no other verse markings within them) are very common. Mapping any arbitrary versification to any other is NICE, but NOT NECESSARY. Displaying the text as the translators intended is NECESSARY. If you can do both, do it. If you cannot, at least display the versification of the Bible translation as the translator intended. I am fully aware of the changes in architecture and code adapting to the realities I perceive imply. At this point, I'm not sure if modifying the Sword engine or rewriting it would be easier. Either way, it is a lot of work. It is my understanding that JSword is a bit better than Sword in this regard, in that it doesn't assume fixed versifications. As far as volunteering for pumpkin holder for versifications, I nominate Arnaud. (I already bit off more than I can chew by myself. Sorry.) On 2/19/24 14:23, Troy A. Griffitts wrote: Dear all, These comments are a mix of background, history, and thoughts: 1) VERSIFICATION (v11n): Variation between reference systems sucks. Until you get into the weeds of the details, it is normal to assume the problems are not complex. SWORD tries to implement a simple 90% solution. SWORD and JSword support defined abstract versification schemes with 3 simple dimensions: [bookid : chapterMax][chapterNumber : verseMax][verseNumber : verseEntry] Conceptually we also operate on these assumptions (I've skimmed the proposal by Arnaud which differs here, but I haven't given it the thought it deserves to comment yet): that book order is defined in the v11n system; that chapter and verse numbers are numeric and begin at 1 and increase to verseMax. We also allocate a special slot '0' for: Module Introduction; Testament Introduction; Book Introduction; and Chapter Introduction (e.g., Matt.0.0 can hold an introduction to Matthew). Those who have been exposed to many Bibles will immediately think of places these assumption fall short. But for >90% of our Bibles, these assumption hold true, and these assumption make many aspects of our work much simpler (abstract parsing of verse lists and ranges, bookmark ordering, etc.). Historically, SWORD previously supported dynamic, per module, versification, with a 3 phase lookup: index file .bks[book number] = book offset in next index; index file .cps[book offset + chapter number] = chapter offset in next index; index file .vss[chapter offset + verse number] = verse offset and entry size in data file. 20 years or so, we made the decision to begin the hard work to understand versification systems within Bibles so we could begin to map them appropriately. This let us remove the .bks, and .cps index files and store that data in versification system definitions, leaving only the final .vss index file which gave the offsets and entry sizes into the data file. Caring about versifications was a decision we made. Our previous design let any Bible decide how many books, how many chapters, and how many verses each chapter contained.
Re: [sword-devel] Catholic versification / inter-versification mappings
Dear all, These comments are a mix of background, history, and thoughts: 1) VERSIFICATION (v11n): Variation between reference systems sucks. Until you get into the weeds of the details, it is normal to assume the problems are not complex. SWORD tries to implement a simple 90% solution. SWORD and JSword support defined abstract versification schemes with 3 simple dimensions: [bookid : chapterMax][chapterNumber : verseMax][verseNumber : verseEntry] Conceptually we also operate on these assumptions (I've skimmed the proposal by Arnaud which differs here, but I haven't given it the thought it deserves to comment yet): that book order is defined in the v11n system; that chapter and verse numbers are numeric and begin at 1 and increase to verseMax. We also allocate a special slot '0' for: Module Introduction; Testament Introduction; Book Introduction; and Chapter Introduction (e.g., Matt.0.0 can hold an introduction to Matthew). Those who have been exposed to many Bibles will immediately think of places these assumption fall short. But for >90% of our Bibles, these assumption hold true, and these assumption make many aspects of our work much simpler (abstract parsing of verse lists and ranges, bookmark ordering, etc.). Historically, SWORD previously supported dynamic, per module, versification, with a 3 phase lookup: index file .bks[book number] = book offset in next index; index file .cps[book offset + chapter number] = chapter offset in next index; index file .vss[chapter offset + verse number] = verse offset and entry size in data file. 20 years or so, we made the decision to begin the hard work to understand versification systems within Bibles so we could begin to map them appropriately. This let us remove the .bks, and .cps index files and store that data in versification system definitions, leaving only the final .vss index file which gave the offsets and entry sizes into the data file. Caring about versifications was a decision we made. Our previous design let any Bible decide how many books, how many chapters, and how many verses each chapter contained. This had its merits because any new versification could be defined in each module without anyone caring what it was. But the drawback was the same: any Bible could decide how many books, how many chapters, and how many verses without anyone knowing why or what they were. Some have pushed for dynamic definitions of v11n systems again, and I understand why. I am in favor of moving forward with a hybrid approach: a set of defined versification systems, which a module will still need to choose from, to which it most closely adheres, + the ability for that module to specify its variation. Toward 98%: We have tried to work around the cons of this simple design and approach 100% support by accounting for the most common types of problems, e.g. * The engine allows common verse suffixes (e.g. Matt.2.7b); * The engine skips verses in a Bible which are not present-- this allows us to create v11n schemes which are a superset of n number of closely related v11n schemes, knowing that the engine will skip over the verses that are not present in the module; We also have tools which print out missing verses which has proven a good QA check for our modules team. * When we run across a Bible which adds an odd verse here or there or an out of order verse, our workaround has been to append these to end of the verse just before where they should appear, so the text flows the same as the printed Bible, and we include for the reader an inline visual separator and marker showing the publisher's verse number. These work arounds get us pretty close to being able to support 98% of our Bibles exactly as the publisher wishes, and the remaining 2% is supported "well enough" for no complaints by publishers. Could we build a system which allowed out of order verses, or which allowed any scheme a Bible wished to follow? Sure, but the added complexity for various tasks increases quite a bit for some of these allowances-- e.g., think index math for book chapter verse when we cannot assume numeric sequence; think abstract ordering of bookmarks not tied to any specific Bible, search results across Bibles, etc. Our vision with v11n definitions is that they will be a few as possible allowing us to map between them most easily; and as many as necessary to allow us to represent well enough a published work. Chris Little previously was our versification pumpkin holder and did some amazing work researching all this material. As a demonstration of his thorough work and an example of the difficulties with v11n, see his work on just the LXX tradition: https://www.crosswire.org/svn/sword-tools/trunk/versification/lxx_v11ns/ Chris has left our community after many years of volunteering massive time and effort. We haven't had anyone step up who is willing to commit the time and