In His Name, the Most High
English
What has changed
daftar is currently running v0.41.0. If something is not working, telling us this number helps a great deal — get in touch.
Your daftar sign-in can now open other toolsv0.41.0running now
daftar can now vouch for who you are to another tool we run — the first is Fihrist, the research toolkit. When you open Fihrist and it asks you to sign in, it sends you to daftar, which confirms who you are and sends you straight back; if you were not signed in to daftar, you sign in as usual and a card on your welcome page takes you back. daftar only says who you are — whether you may use the other tool is decided there, by its own list of members. Nothing changes about signing in to daftar itself, and nothing here is switched on until Fihrist is ready. Nothing to do at your end.
Record a lesson inside daftarv0.40.0
You can now record a lesson without leaving daftar: fill in the teacher and subject, press Continue, then Start. A beep sounds and a red indicator with the time stays on screen while it records, so everyone in the room knows. Pause and resume as you need — each part is kept, and the gap between them is remembered rather than hidden. Everything is saved on your phone every few seconds as you go, so a crash costs seconds, not the lesson. Press Stop and the recording is sent; if the connection is gone, it waits on your phone and the list of recordings waiting to be uploaded offers “Send now”. On an iPhone, keep the screen on and daftar in front — the phone stops the microphone otherwise; the page says so, and the phone’s own recorder remains an option. Nothing to do at your end unless you want to try it.
Recording on your phone and uploading later — with a reminder that never gives upv0.39.0
The upload form now asks how the recording will get here. If you will record with your phone’s own recorder and upload afterwards, daftar saves the lesson’s details straight away and keeps a place for the file. A bar at the top of every page then says how many recordings are waiting, and opens a list where you can upload each one — or say “I did not record this” if it never happened. If a recording is still waiting after a day, you get one email in all three languages listing what is waiting; if it is still waiting a week later, another, and after that once a month — sent late Thursday night, never in the middle of your studies. Nothing is ever cancelled for you. Nothing to do at your end unless you have a recording waiting.
A security check on every change, from now onv0.38.1
Every change to daftar now gets a security review before it goes out — who can reach what, what a person could type or upload, and what an error message might reveal. The first one caught real problems the same day the recordings feature launched, so it is now a standing rule rather than a good intention. Nothing to do at your end.
The upload form now looks like the rest of daftarv0.38.0
The form for uploading a recording was built in a hurry and did not match the others — smaller boxes, a bare browser file button, a cramped layout. It now uses the same fields, spacing and buttons as the request-access and sign-in pages, with a proper file picker that shows the name and size of what you chose. Behind the scenes, every form in daftar now draws its controls from one shared set, and a check refuses any new form that does not — so they cannot drift apart again. Nothing to do at your end.
A security check of the new upload feature, the same evening it went livev0.37.1
We now do a security review of every change before it ships, and the first one found three things in the upload feature released a few hours earlier. The most important: a signed-in contributor who somehow knew the internal ids of another contributor’s recording could have read it, or interfered with it while it was still uploading. Nothing in the app reveals those ids, and there is no sign anyone tried — but the door is now locked properly: an upload can only be touched by the person it belongs to. Two smaller things were tidied at the same time, including one that predates the recordings entirely. Nothing to do at your end.
You can now upload a recording of a lessonv0.37.0
Record a lesson with your phone’s own recorder, then upload the file on your recordings page — say who taught it, what it was and when, and it goes up. Large files carry on by themselves if the connection drops; the page tells you when that is happening. Once we have checked the file it is marked ready, and it is yours to download from then on, exactly as you sent it. Recording from inside this page is next. If a file cannot be read, the page says so in plain words rather than quietly accepting it. Nothing to do at your end unless you have a lesson to upload.
Room for lessons and recordings, and a save that had quietly never happenedv0.36.0
Behind the scenes the database now has a place for teachers, lessons, recordings, transcripts and translations — empty for now, so that when the first recording arrives every record it produces can say exactly which lesson it came from. That link cannot be added afterwards, which is why it goes in first. One fix you will not notice but should know about: when you closed the tab mid-sentence, the form tried to save one last time on the way out — and that particular save had never reached us. The regular autosave a couple of seconds after you stop typing always worked, so at most a few seconds of typing were ever at risk. It saves properly now. Nothing to do at your end.
Homework before the recorder: what phones will and will not let a web page dov0.35.2
Before building the button that records a lesson, we wrote down — with sources — what each kind of phone actually allows. The one that matters most: an iPhone stops a web page recording the moment you lock the screen or switch apps, and no web page can change that. So on an iPhone the recorder will work while the screen is on and the app is in front, and for a long lesson you will also be able to record with the phone’s own recorder and upload the file afterwards. Android phones keep recording in the background. Nothing visible changes yet; this is the groundwork. Nothing to do at your end.
Our own check said the site was down, on the day it started speaking three languagesv0.35.1
Straight after the last release went out, the check we run to confirm a release worked said the home page had not rendered. It had — in all three languages. The check was asking for the page the way it was addressed before the languages existed, and getting a polite “this way” instead of a page. It now asks for each language by name and confirms it gets that language back, which is a stronger question than it asked before. Nothing to do at your end.
The subject names are in your script nowv0.35.0
Until today the dropdown for a record’s discipline said usul, fiqh, kalam in Latin letters whatever language you were reading in, and so did the list of your records. They are Arabic words, so that was odd. If you read in Arabic or Farsi you now see them written as you would write them; in English they are capitalised, with a word of explanation beside the less obvious ones. What is saved with the record has not changed at all — the same short label as before, so a record written in Farsi and one written in Arabic about the same subject are still, to the dataset, about the same subject. Only what you see is different. Nothing to do at your end.
Some of it was still in English, and our own check said it was notv0.34.0
If you were writing a record in Farsi or Arabic, a handful of things on that page were still in English — the box about a question touching somebody’s own obligation, the button that sends a record for review, the ones that add a citation or a row. They are in your language now, along with the reviewer’s panel, the names of the seven answer patterns, and the small print beside your passkeys. The honest version of what happened is worse than a missed translation, and worth telling. Those words had already been translated. They were written down, read back and corrected months ago, and they were sitting in the site’s own dictionary the whole time. The pages just never looked them up. So the work was done and the reader never saw it. What we are less comfortable about is that we had a check for exactly this, and it passed every single day. It was asking whether each English sentence had a translation somewhere — and every one of them did. It never asked the question that mattered, which is whether the page was actually using it. The test we had written to prove that check worked was built on the same assumption, so it agreed. We have changed what it asks, and we have added the check we should have had from the start: one that loads each page in Farsi and in Arabic and simply looks at what is on the screen. If it finds English where English should not be, it says so. It also refuses to report a clean result unless it can first prove it is still able to see anything at all, because a check that has quietly stopped working looks identical to one with nothing to find. It earned its place immediately. On its first run it found a sentence on the Farsi terms page that had been damaged by a copy-and-paste — half of it repeated, and a link showing as raw brackets instead of a link. That is on a page about what we promise you, so we are glad it is fixed. One thing we removed rather than translated: the words “written by” in front of the author’s address at the top of a record. Beside a status that was already in your language it read as clutter, so it is gone in all three. Nothing to do at your end.
daftar speaks Farsi and Arabic now — all of it, not some of itv0.33.0
Every page, every button, every form, every error message, and both of the emails we send you. You can change language whenever you like from the control at the top of any page, and it stays changed. We should be straight about why this took until now. We were asking people who study and teach in Farsi and Arabic to do careful work in English, on a site that was entirely in English, while telling everyone else that the thing we are building would speak to them in their own language. That was the wrong way round, and we have fixed it. None of it is machine-translated. Every line was written and then read back, word by word, by someone who speaks the language — including the awkward ones, like the difference between a record sent back for changes and one not accepted. Those are different things, and somebody who reads the first as the second would reasonably stop contributing. The emails go out in all three languages at once, every time. We thought about asking whoever invites you to pick one, and decided against it: they would be guessing, and a guess about which language somebody reads is the kind of thing that is wrong quietly. You choose. This page stays in English. It is our log of what we have done, and it is the one place we would rather be plain than polished. Nothing to do at your end.
Two pages we would rather have than not, before they are finishedv0.32.0
daftar now has a privacy page and a terms page, in English, Farsi and Arabic. Neither is the finished document — those are being written with our legal advisers, so that they hold up where you actually are and not only where our servers are. We would rather publish them late and correct than early and vague. In the meantime both pages say what is already true: we use no tracking of any kind, we hold no password for you because there is none, your recordings and records are yours and we are their custodian, your name stays on your work, and there is no public sign-up and never will be. None of that is waiting on a lawyer. They are things we have already said out loud to everyone we have spoken to, and the document will record them rather than introduce them. If you want to know what we hold about you, or want it removed, you do not have to wait for a document — ask, and a person will answer. Nothing to do at your end.
Two pages that were bare now have something behind themv0.31.0
The page you land on after signing in, and this one, now have a background image like the others do. The first is an open, empty notebook under a lamp — daftar is the word for that book, and a blank page waiting is more or less what the page is asking of you. Neither picture contains any writing, and that was the first rule we set: a machine drawing Arabic produces something that looks like script and is not, and that is not a thing to put behind text on a site read by people who would spot it immediately. Nothing to do at your end.
The page you land on now tells you what needs youv0.30.0
When you sign in, you now see what needs you — the records you are still writing, and the ones a reviewer has sent back — and, separately and more quietly, the ones that are with a reviewer. We split those on purpose: a record waiting on somebody else is not something you can do anything about, and putting it at the top of your page would just teach you to check it. It is still shown, so that an empty list means “nothing to do” rather than “where did they go”. Once you have a passkey, that box folds itself away. It is a prompt until you have one and a settings list afterwards, and it was pushing your actual work down the page. Click the heading to open it. There are also two greyed-out panels for things that do not exist yet: recording a lesson, and getting the transcript back. They say “designed, not built yet” and they carry no date — this page once said “coming in the next release” for several releases in a row, which is worse than saying nothing at all. You will hear about them when they work. Nothing to do at your end.
A page that says what taking part actually involvesv0.29.0
There is a new page — Roles, in the footer beside About and Passkeys, and in English, Farsi and Arabic. It sets out who we are looking for, what the work actually is week to week, who owns what, and what we are not asking for. It exists so that you can decide before anybody asks you to, rather than after. The part worth repeating: your recording, your transcript and the questions and answers you write are yours — downloadable whenever you like, for as long as daftar exists, whether or not you are still contributing. Your name stays on your work. The one thing that is ours is the assembled dataset. The Farsi and Arabic say plainly at the top that they have not been reviewed yet. That notice comes off when they have been. And one thing we got wrong and fixed: the two role descriptions were side by side, which squeezed each one to about thirty characters a line — for the very text you would read to work out whether this is something you could do. They are stacked now. It is the same mistake we fixed on the record form last week, made again in a different file. Nothing to do at your end.
Nothing changed in the tool — we wrote down what we are asking of a madressahv0.28.0
Nothing in daftar itself is different this time, so there is nothing new for you to try. What we did was write down, properly and in one place, what this project is, the limits built into it, what one finished record actually looks like, who owns what, and what we would like to ask of a madressah — in English, Farsi and Arabic. It is the conversation we have been having in person, written so that it can be read without us in the room. The part worth repeating here, because it applies to you and not only to an institution: your recording, your transcript and the questions and answers you write are yours. They stay available to you and downloadable for as long as daftar exists, and your name stays on your work. The one thing we own is the assembled dataset. We also made the machine check something a person cannot: a slide twenty pixels too tall looks perfect on screen and quietly prints as two pages. Ours did, on the slide showing a real record. It does not any more. Nothing to do at your end.
You can write a record now — and it saves as you gov0.27.0
This is the one that makes daftar useful. You can write a record. A question, an answer, and the sources the answer rests on — with the citations as proper fields rather than one long line you have to punctuate correctly. It saves while you type. Not when you press something, and not when you remember — on its own, every few seconds, and again if you close the tab mid-sentence. The little note at the top only says “saved” once our server has actually confirmed it, never merely because your browser sent it. If it cannot save, it says so plainly rather than sitting there looking calm. The seven kinds of answer are explained in the form, next to where you choose one, with the mistake people usually make with each. Three of them we are short of, and it says so when you pick them. It stops you where a rule is a rule, and only asks where judgement is yours. A question about someone’s own obligation must send them to their marjaʿ — that one is refused outright. Whether an answer really does refer them onward, rather than merely using the word, is something only a person can read, so it asks instead of pretending to know. There are seventeen finished records to look at. Real ones, written out properly — every kind of answer, every discipline, including one held back because a page number still needs checking, and one that was refused, with the reviewer’s reasons written out in full. Reading what review turns away teaches the format faster than any description of it, so that one is there on purpose. They also showed us something missing: finished records were not listed anywhere. You could open one if you knew where it was, and find it no other way. Reviewers now see an Already reviewed list. And the form is laid out down the page now, one field per line in groups, instead of two columns. The columns looked neater and read worse — a label with a note under it pushed its neighbour out of line, so the rows sat crooked. One thing we caught before you did: that rule could be reached two ways, and only one of them worked. It is possible to mark a record as touching a personal obligation, and also to choose “boundary” as the answer type — which mean the same thing. Doing the second without the first slipped past the check entirely. Both now trigger it. Nothing to do at your end.
Passkeys now work underneath — there is just nothing to press yetv0.26.0
The machinery for signing in with your face or fingerprint is now in place and working. What is missing is the button, which comes next. So there is still nothing for you to do. One decision worth telling you about, because it affects you and cannot be undone later. A passkey is tied to this address and only this address — that is what makes it impossible to use on a convincing fake of our site. It also means a passkey you add here will not work on our test site, and vice versa. That is the protection working, not a fault. We have also chosen to always ask for your face, fingerprint or PIN, rather than accepting a simple tap. Some sites allow the tap, because it is quicker. For something that records who wrote what, we would rather know it was you. And a small thing we caught while writing the safety checks: a web address ending in our own name is not necessarily ours. evil-tpilmflow.com ends withtpilmflow.com, and the obvious way to write that check would have accepted it. Ours does not, and there is now a test named after that exact mistake. Nothing to do at your end.
We closed a door that was open, unused, and should not have been therev0.25.0
A security fix. Nothing you can see changes, and there is nothing you need to do. The software this site is built on had a serious flaw published against it, in a part that resizes images on demand. We do not use that part — our images are prepared in advance rather than resized as you visit — but it was switched on by default, and anyone on the internet could reach it. So we did two separate things, on purpose. We turned that feature off, since nothing here needed it. And we updated to the version where the flaw is fixed. Either one would have been enough; doing both means neither has to be relied upon. The uncomfortable part is how we found it. We were adding something unrelated, and the tool that installs it happened to print a security summary on its way past. Nothing we run asks that question — so it was not missed through carelessness, it simply was never checked. We are fixing that too, because a thing nobody checks is a thing nobody knows. Nothing to do at your end.
More passkey plumbing, and a check that stops one door opening anotherv0.24.0
Nothing you can see yet. This release adds the part that remembers a sign-in attempt while it is happening — a short-lived note that expires after five minutes and can only be used once. The interesting bit is a check that keeps two things apart. Adding a passkey and signing in with one look similar from the outside, and if a system muddles them, somebody could use the adding step to get themselves signed in as you. Ours refuses that, and refuses it before it checks anything else, so the attempt is recorded for what it actually was rather than filed away as “expired”. We also caught ourselves being untruthful. When one of these attempts is turned away, we show a short message. The first version said the attempt had “expired or was already used” — true in two cases out of five, and simply wrong in the other three. It now says only that the attempt is no longer valid, which has the advantage of being true every time. Nothing to do at your end.
A page explaining passkeys, before anyone asks you to use onev0.23.0
There is now a page that explains what a passkey is — in plain language, with no jargon, and in English, Farsi and Arabic. It is public, so you can read it without signing in, and you can send it to anyone. We have deliberately put it up before the feature itself. An explanation that arrives after people have already been confused has missed the point of being an explanation. So there is nothing to do with it yet; it is simply there, so that when you are eventually asked whether you would like a passkey, you have already had the chance to find out what one is. It answers the questions people actually ask. Do I need to install anything — no. Do I need my phone — only if you are on a computer that does not have your passkey, and then it is a barcode and your camera. What if my computer is older, or has no Bluetooth, or runs Linux — there is a section for that, and the short answer is that you are never stuck. What if I lose my phone — your email link keeps working, always, and that is not going to change. There is no minimum specification for using daftar, and there will not be one — if your computer can open this page, it can sign you in. It also says plainly where not to add one: a shared or borrowed computer is the wrong place, and we would rather say so than have someone find out. The Farsi and the Arabic have been read through and corrected properly before going up, rather than after. Almost every correction was about register — the draft kept reaching for the bookish word where the everyday one belonged. In both languages the sentence written to stop you worrying used a word closer to “you are in real trouble”, which is not what we meant at all. And in Arabic we had written that a security key needs no radio — perfectly grammatical, except that in Arabic that word means the thing that plays music. If anything still reads oddly to you, please tell us; you will be right, and we would like to know. One thing we fixed on the way: the check that keeps our own name written correctly everywhere could not see the file this page is written in, and would have gone on reporting that all was well. It also pointed at the wrong line when it did find something. Both are mended, and we proved it by breaking the page on purpose and watching the check catch it — a check you have only ever seen pass is not one you have seen work. Nothing to do at your end.
Groundwork for signing in with your face or fingerprintv0.22.0
We have started building passkeys — a way to sign in using whatever already unlocks your phone or laptop, instead of waiting for an email each time. Nothing is visible yet; this release is the plumbing underneath. Two things worth saying now, because we would rather you knew than be surprised. You will not need to install anything. Nearly every phone and laptop already has this built in. If you sign in on a computer that does not have your passkey, it shows a square barcode, you point your phone’s camera at it, and that is the whole of it. When we switch this on there will be a page explaining it properly, in plain language, before you have to decide anything. Your email link will keep working, always. A passkey is an addition, never a replacement, and it is how you get back in if you lose the device. You are free to ignore passkeys entirely and nothing about your account is worse for it. One note for the curious: writing the very first part of this, we came within a hair of a rule that would have locked out every single person using an iPhone or a Mac — the obvious way to write one particular check is wrong in a way that only shows up with real people and real devices. We caught it, and there is now a test that fails if anyone re-introduces it.
A power cut showed us that our sign-in form could leak who our contributors arev0.21.2
A battery backup failed this evening and took our mail server off the network for about forty minutes. That is a mundane sort of problem, and it revealed one that is not. Our sign-in page is built to give you exactly the same answer whichever address you type, on purpose. This place is invite-only, and if the form said “no such account” for one address and “check your email” for another, anyone could use it to work out who is on the list. That is not information we are willing to hand out about the people who contribute here. With the mail server unreachable, that stopped being true. A real member’s address produced an error page; an address we had never seen produced the usual friendly message. For about forty minutes the form answered the one question it is designed never to answer. It is fixed, and the fix is covered by tests that fail if anyone undoes it. The uncomfortable part is why we had not caught it sooner: every test we had written for this checked the two answers matched while our mail was working. None of them could see what happened when it was not. We have written the ones that can. Nothing to do at your end, and nothing was lost — sign-in links were not sent during the outage, so if you tried and nothing arrived, please just ask again.
If our email lands in your junk folder, please tell your mail appv0.21.1
Our first invitation to a Gmail address passed every security check a mail provider makes — and went to the spam folder anyway. Nothing was wrong with it. We are simply a new sender: the address our mail comes from has no history, and filters are cautious about a short message whose main content is a link, which is a shape they see in genuine scams every day. So the sign-in and request pages now say plainly to check your junk folder. And if you find one of our messages there, marking it “not spam” genuinely helps — a person moving a message to their inbox carries more weight with a mail provider than anything we can do at our end, and it is the fastest way to stop it happening to the next person. It should settle down as we send more.
The requests people send us can now actually be acted onv0.21.0
Until today a request to take part was recorded and then sat there — there was no page that listed them and no way to say yes or no. There is now a queue, oldest first, so nobody who has waited longest gets overtaken. An administrator can invite someone straight from their request, or set it aside with a note for our own record. Two deliberate choices worth stating. If you wrote to us by email or telephone rather than using the form, we can now add you to the same queue by hand — a name that reaches an inbox and not the queue is exactly the name that gets lost. And if a request is set aside, nothing is sent to you: the form already says we are not onboarding widely, and a rejection letter to someone who volunteered to help would be a worse thing to receive than silence.
Invitations exist now — which, awkwardly, they did not beforev0.20.0
daftar has always been invite only, and until today there was no way to actually issue an invitation. The machinery for accepting one was complete and tested; the part that sends one had never been built, so the only account in existence was made by hand from the command line. There is now a proper administration page for inviting someone, choosing what they can do, recording whether they teach or study, and withdrawing an invitation that has not been used. If you asked for access, this is the thing that was missing. Two smaller repairs came with it: our emails now come from an address that can actually receive a reply — sign-in links were being sent from a domain with no mailbox behind it, so anyone who hit Reply was writing into nothing — and we have started keeping a record of administrative actions, so that “I never got an email” can be answered by looking rather than guessing.
Making sure a sign-in email that fails tells us it failedv0.19.5
Yesterday’s fix got mail moving again. This one is about what happens when a message does not arrive. Email carries two different sender addresses — the one you see, and a hidden one that decides where a delivery failure gets reported and how strongly the receiving service trusts us. We had configured the hidden one carefully, checked it at start-up, and then never actually attached it to the message. So failures were being reported to an address that could not receive them, and an email that never arrived looked exactly like one that did. Both are fixed, and our tests now check the message itself rather than the setting.
Sign-in emails were not being sentv0.19.4
The first real sign-in email did not go out. Our mail server was offering an encrypted connection it could not actually provide, so nothing was ever delivered — and our own checks had missed it because they never tried to encrypt. Fixed, and the check now does what a real client does.
Checks that were checking the wrong thingv0.19.3
Nothing you will notice: two of our own deployment checks were testing for things that had deliberately changed, and reported a healthy release as broken. Both are fixed.
Refusing to start rather than fail quietlyv0.19.2
daftar now checks two more mail settings when it starts up, and refuses to run without them — rather than starting normally and failing later, when somebody is trying to sign in or ask for access.
A stray box on the request formv0.19.1
A spam-protection field that is supposed to be invisible was showing up above the send button. It is now properly hidden.
You can ask to take partv0.19.0
There is now a form for requesting access, for people who teach or study in a hawza and would like to contribute. Requests are kept and read — but daftar is still invite only, and we are not onboarding widely, so the form says plainly that it may be a long wait.
A way back from the About pagev0.18.1
The name at the top of the About page now takes you home, and there is a link beside the language choices too.
The English caught up with its translationsv0.18.0
Reviewing the Farsi and Arabic turned up a vagueness in the English they were translated from, so the English has been corrected to match them.
The Arabic has been reviewedv0.17.0
Both the Arabic and Farsi About pages have now been read by hawza-trained speakers, so neither carries a draft notice any longer.
The Farsi has been reviewedv0.16.0
A Farsi-speaking hawza reader has been through the Farsi About page, so it no longer carries a draft notice. The Arabic still does, until it has had the same reading.
Safer updatesv0.15.0
Nothing you will see directly: the process that puts a new version of daftar in front of you now refuses to run if the new version was not actually built. Previously that situation looked like a successful update that quietly changed nothing.
IlmFlow is now a linkv0.14.0
Wherever IlmFlow is mentioned on the About page — in any of the three languages — it now links to the project site.
daftar, always lowercasev0.13.0
The name is written lowercase everywhere, as a small stylistic choice — and the build now checks it, so it stays that way.
An About page, in three languagesv0.12.0
There is now a page explaining what this project is for and — just as importantly — what it deliberately will not do, in English, Farsi and Arabic. The Farsi and Arabic are a first draft awaiting review, and each says so at the top in its own language.
The name in both scripts, and a way back after signing outv0.11.0
daftar now carries its name in English and Arabic on every page and in every email. Signing out used to leave you on the sign-in form with no way anywhere else — it now takes you back to the front page and tells you it worked.
A better-looking sign-in emailv0.10.0
The email carrying your sign-in link now has proper formatting. It still shows the full web address in plain text next to the button — deliberately, so you can always check where a link goes before clicking it. We will never send you a link that hides its destination, and neither should anyone claiming to be us.
daftar looks like daftar nowv0.9.0
The colours are the ones tpilmflow.com uses, so this feels like part of the same thing rather than a separate tool. There are photographs on the front and sign-in pages — a seminary courtyard and a lamplit library — and they are small enough to load quickly on a phone.
Teachers, students, reviewers — and nobody marks their own workv0.8.0
Accounts now record whether you teach, study, or both — because in the hawza most people are both at once, and a field that made you choose would be wrong for most of you. Separately from that, reviewing is now its own permission, so somebody can check records without also being able to create accounts. And nobody can approve their own record, including administrators: the whole value of the review step is that a second person opened the book.
Sign-in emails can actually reach youv0.6.0
We found three ways a sign-in email could have failed without anyone noticing — including one where a message that never arrived would have looked exactly like one that did. All three are fixed, and daftar now refuses to start at all if its mail settings are wrong, rather than starting up and quietly failing later.
The right colours, and you can install it on your phonev0.5.0
daftar now uses the Truth Promoters palette, checked for readability rather than chosen by eye. You can also add it to your home screen on a phone or tablet and it opens like an app.
Sign in with a link sent to your emailv0.4.0
No password to choose, forget or reuse. daftar is invite only — there is no sign-up page, and a link only ever arrives after you ask for one.
Somewhere to keep the recordsv0.3.0
The database behind daftar — records, citations, key terms and who reviewed what.
One set of rules for the form and the converterv0.2.0
The vocabularies and checks now live in one file that both the web form and the conversion tool read. Previously each had its own copy, which is how they drift apart and start disagreeing about what a valid record is.
The beginningv0.1.0
The first deployable version. Nothing to see yet.