================ Datatracker PLAN ================ Updated: $Date$ Planned work in rough order =========================== * Polish the htmlization pages, making the style identical with tools.ietf.org. * Wrap Trac in a docker container, in preparation for transition to running the datatracker under Python 3. It doesn't look like Trac will be available for Python 3 until January 2020 at the earliest. * Investigate making RFCs first-class document objects to faciliate being able to model BCPs that represent groups of RFCs properly. Then fix the rfc sync script to update BCP aliases when changes occur changes (a BCP points to new RFC numbers) * Complete the transition to Python 3. This will make it easier to add add support for internationalsed email, and also other i18n enhancements. The code has now been made Python 2 / Python 3 compatible (as of release 6.99.0, 16 July 2019), but has not been deployed under Python 3 yet. The natural point for changing the deployment seems to be just after the next OS upgrade, which we think will make Python 3.6 the default mod_wsgi python. * Review and change the draft submission and handling code to deal with UTF-8 instead of ascii before upload of utf-8 drafts should be permitted. (This may fall out for free with the Python 3 transition). In ietf.submit.views.upload_submission(), for instance, there's a file open() which should be io.open() with the appropriate encoding. In general, there are about 100 file open() which should be reviewed and changed to io.open() unless that's clearly not right. * Add generation of bibxml reference library entries to the datatracker, in order to be able to transition away from the old TCL code. * Add django pagination to selected pages with long lists of results in order return results to the user faster. * Change our web asset handling from the current bower-based (which has been deprecated and gradually more and more unworkable) to a yarn-based setup. Short-time the effect of not having the web asset updates in mkrelease is negigible, but long-term the lack of this means that we may miss security-updates to web assets. * (EXTERNAL BID) Reworked UI and refactored backend for the scretariat meeting scheduling tool. * Add support for internationalised email addresses according to RFC 6531 when sending email (this is not supported by smtplib under Python 2.7, so will need python 3.x). * Test suite improvements. Reduce and seek to eliminate causes of differences in code coverage between runs in the same and in different environments. * Rework email sending so that all emails sent by the datatracker, except for logging and failure emails sent to ADMINS, are captured as a Message object. * Rework email subjects for messages relating to documents to consistently start with the document name. * Revisit RFC author ordering and author information based on rfc-index.xml * Revisit photo uploads: Add photo upload for people with roles. Add photos to wg pages and group overview pages. * Transition to Django 2.x (depends on Python 3.x). Security updates to Django 1.11 will cease around April 2020. * Revisit floorplans: Add room coordinate input tool (javascript). Add NOC object annotation possibilities. Add support for break areas on multiple floors (probably means modelling break areas the same way as other sessions, and removing the break_area char_field on Meeting. * Add the ability to volunteer for Nomcom when logged in to the datatracker. * Transition to using timezone-aware timestamps. Migrate database timestamps to be timezone-aware. This means converting all timestamps to PST8PDT, except for meeting-related timestamps, which use the meeting.time_zone. Once done, set USE_TZ to True in settings. * Add support for document shepherding reports, possibly re-using or generalising some of the review plumbing. Check with IESG for details. * Transition to PostgreSQL. This will make it easier to start using timezone-aware timestamps throughout the code, which will make it easy to present localized times on web-pages. It will also provide additional tools for performance analysis * Performance analysis of database table and index setup * Refactor Document and types into Document subclasses, with conditional code in views and utilities moved into models and overridden models where handling differs between document types. * When draft XML source is available, take references from that instead of extracting them from the text version. * Integrate better author data for RFCs from either the RFC Editor or from running the author extraction script on newly published RFCs. * Add an FK to person from author, and use that instead of the unknown-email-* addresses to connect documents to authors when no email address is available. (we still need to keep author email addresses in order to know which address to associate with a given authorship, when we have an address). * Review places where we display persons' names but link to email addresses. Some of those probably should link to the person's profile page, instead. Notes =========================== * Small nomcom refactoring: associate Feedback records with Person instead of User. * Consistency fix: in settings.py, name directory settings consistently with ..._DIR; reserving ..._PATH for settings with PATH semantics (':'-separated lists of directories) * For documents with Yang modules, add links to the extracted modules (possibly in multiple formats -- pyang can generate a large number of alternative formats) * When we get to the point where we can produce pdf from xml+media, remove the pdf upload possibility -- it's a definite attack vector. Possibly start scanning pdf files for /JS and /JavaScript (and workarounds to hide Javascript) * Add one or more API description pages with examples of common queries * Add role lists and additional relevant information to the personal profile pages * Increase the requirements on valid email addresses in draft submissions -- contacting authors is hard if they don't provide an email address. * Make it possible to let time run backwards in the database (creation timestamps, universal changelog) * Change slugs which were limited by the former 8-character slug length limit to be more readable (needs both code changes and migrations). * Additional charts, various statistics views. * For graphing and mapping various numbers relative to country population, integrate population data from the world bank, http://databank.worldbank.org/data/reports.aspx?source=2&series=SP.POP.TOTL&country= (or some other source, if a better is found). WorldOmeters has running counters, but ask for a lot of money for even one. * Notable New Django Features in 1.8: + New data types: UUIDField, DurationField + Query Expressions (enhanced expressivity, conditionals, database functions) + Improved TestCase fixture loading, additional .setUpTestData() method. - Support for alternative template engines - Security enhancements (request/response cycle middleware) - PostgreSQL-specific fields * Notable New Django Features in 1.9: + New on_commit() hook, for post-commit actions, for instance sending email + Password validation option with pluggable validators + Running tests in parallel + DateTimeFields can be queried with date - Permission mixins for class-based views - New admin style * WG document state slug renaming. adopt-wg -> adopted-by-wg c-adopt -> candidate-for-wg-adoption chair-w -> waiting-for-chair-goahead dead -> dead-wg-draft info -> adopted-for-wg-info-only parked sub-pub -> submitted-for-publication wg-cand -> wg-candidate wg-doc -> wg-document wg-lc -> wg-last-call writeupw -> waiting-for-shepherd-writeup * Clean up and normalise the use of canceled / cancelled. User-visible strings should already be all CANCELLED except when manually entered. * DjangoCon Europe 2017 Notes: - Add tracing of Django Query methods from code through templates to the sql_queries list provided by 'django.template.context_processors.debug' (DONE) - Consider rewriting user switches using feature flags, for instance with gargoyle. - There is now a Django-REST-Framework add-on app which makes it easier to do something similar with DRF as with Tastypie. As Tastypie is not being actively maintained, and DRF seems to have better performance, consider building /api/v2 using DRF and drf-schema-adapter / drf-auto-endpoint. - Consider adding JWT (RFC7519) support for /api/v1 /api/v2, to generate an access token from a login, and use that instead of session support for access control to access limited endpoints. - Once we're on Django 3.5, start using static type annotations to improve early discovery of incorrect function/method usage. There are add-on files avaliable for Django which provide type annotation for Django functions and methods, and the Python stdlib is type annotated starting with Python 3.5. Check static typing violations with 'mypy'. - Consider providing a user-selectable option to import photos from gravatar. - Consider using django-12factor to apply part of the 12-factor app philosophy: https://12factor.net/