Update:Remora Meeting Notepad/Archive Apr 2008-Jun 2008
- 1 June 25th
- 2 June 18th
- 3 June 11th
- 4 June 4th
- 5 May 28th
- 6 May 21st
- 7 May 14th
- 8 May 7th
- 9 April 30th
- 10 April 23rd
- 11 April 16th
- 12 April 9th
- 13 April 2nd (Post-Mortem)
- AMO 3.4.5 (maintenance release) status - Code Freeze: 7/2, Push: 7/8
- wenzel's status:
- bug 432043 "No way to see experimental versions of public add-on": Done, waiting for review (fligtar).
- bug 436810 "Search results only show 1 review": No code change necessary. The cron jobs were allegedly put on preview, but I reopened bug 438832 to have IT double-check as the problem is still present. Question: How about we have the reviews and ratings automatically adapted every time somebody files/deletes a review? That'd keep the counts up to date and still save us the real-time counting queries on every page request. If you like it, somebody please file a bug.
- bug 441463 "Add link to new reviews guideline to quick review box": Patch prepared, waiting on review (clouserw).
- wenzel's status:
- AMO 3.5 (Dev Tools) status
- API - next steps?
- Will work with IT to do a AMO code rollback and watch the NS & App server loads for 15 min and then roll forward. It will help isolate the load issues.
- Debrief on Fx 3 launch (quick review of lessons and points to remember)
- AMO 3.4.5 status - keep current schedule or re-evaluate?
- 3.4.3 code release scheduled for Thu (6/13) evening
- Anyway to put the AMO version/SVN string on the site? bug 425085. Ryan to file bug on packaging
- Dev Tools: Preview panel has been re-written (going into AMO 3.5)
- Dev Tools left to do: Submit & Update process
- 3.4.5 Code Freeze: 6/26 Push: 7/1
- No Notes
- AMO v3.4.3: NOTE new release date of Jun 5th
- AMO v3.5: Will add to preview so QA can test. After that, we'll look to put it in parallel on public AMO
- AMO v3.4.3
- Code Freeze: Mon, June 2nd @ 11:59pm PST
- QA & Verification: Tue, Wed
- Push: Thu, Jun 5, 7:00pm
- After fixing bug 425481, there's no way to get a complete list of just the extension type now . Do we want that? If so, where would that go UI-wise? (Update: Somebody filed bug 434328 about this). --wenzel
- When auto-approving reviews, what do we want the process to be for marking comments to be moderated (bug 434048)? Spam/inappropriate/other (+ comment box)? --wenzel
- Action: For 3.4.3, we'll just add the "report a review" and we'll discuss and design for upcoming releases.
- Timing of 3.4.4 vs Fx 3 --laura
- Search discussion  --laura
- Review of draft AMO Operational Plan --basil
- review categorization mockup (regarding bug 432665 and it's dependencies)
- AMO 3.5 - Shooting now for Thu 5/29 to be put on preview
- AMO 3.4.2 status - we are good for a Thu push
- Wil to file the IT bug & resolve bug 430877
- How are the regular cycles going?
- Setting cut-off dates (make it the same time every two week cycle for code freeze)
- We need to write new tests
- We need keep doing the UI work in pre-cycle work so that the 2-week cycle is used for coding and test
- AMO 3.4.3 triage
- AMO 3.5 - Shoot to get new Dev Tools by next Thu 5/22 (earliest)
- Glimpse preview (VPN required)
- Would be nice to pick up bug 432160 and bug 431576 (RSS and category layout issues) for 3.4.2, if possible -- stephend
- Password Reset regression - can we do a quick fix before 3.4.2 comes out for bug 432584 -- basil
- Code Freeze: Mon, May 12th @ 11:59pm PST
- QA & Verification: Tue, Wed
- Push: Thu, May 15, afternoon
- [Clouserw] Map out a plan for how to deal separate L10N releases from mainline AMO releases so that we can continue on the 2-3 week release pace.
- Code Freeze: Tue, April 29 @ 5pm PST
- QA & Verification: Tue, Wed
- Push: Thu, May 1, afternoon
- What to do about L10N changes?
- Apparently, the "missing" search button confuses people, in combination with Enter on the category dropdown not submitting the search request. bug 430128. Should we deal with this in 3.4.1? Enabling the button even with JS enabled would be an easy fix. What about its CSS style? --wenzel. Targeted bug 430128 for 3.4.2.
- What about adding IceWeasel and IceApe to the browser recognition (bug 429924)? Do we fear needing to add other Firefox derivates also? Is there a problem when we support these products here but not elsewhere on the page? --wenzel. Minor fixed applied for 3.4.1. Not fully supporting these browsers.
Proposed schedule from fligtar:
4/7 - Edit: Manage Authors, Translation Box
4/14 - Edit: Descriptions, Edit: Categories
- 4/21 - Edit: Properties, Edit: Versions and Files
- 4/28 - Edit: Manage Preview Screenshots
- 5/5 - Merge with trunk and push live for optional "Try out our new beta add-on editor" testing (If things go more quickly, may be able to do this push on 5/1 with 3.4.1 stuff?), Submit New Add-on
- 5/12 - Update Add-on, Nominate/Change Status, General fitting in & cleanup
- 5/19 - l10n, preview.amo testing
- 5/26 - additional l10n, preview.amo testing?
- 5/29 - launch as default & only dev cp
- Increase default add-on count on listing pages to 20 (from 10?) --wenzel
- Improving usability by creating a more straightforward work flow for the most basic use cases? (Should we make a bug to track this, or a wiki page that will state the action items that have to happen?) --wenzel
- getting rid of hybrid pages between themes/appearance, search engines/search tools and dicts+langpacks/language extensions? Instead, give them their own landing pages, and link to the related extensions categories from there for those who are interested.
- I suggest getting rid of the categories list on the LHS on the front page, instead providing mainly four icons (or so) to reach each of the four spots: extensions (= the current home page), themes (the new theme browser), search engines (list or landing page), language-related (landing page).
- Additionally, we should of course keep recommending some (2 or 3) add-ons on the front page, and promoting the newest, updated and recommended lists along with their respective RSS feeds.
AMO 3.4 milestones
- 3.4.1 (Cycle 1) - April 29th is the target date for code freeze and we'll push on May 1st.
- 3.4.2 (Cycle 2) - Approx May 2 - May 14
- 3.4.3 (Cycle 3) - Approx May 15 - May 30
- Cycle 1 priorities:
- compatibility ranges, IE7 download issue
- themes/appearance de-hybridization
- search engines & search tools de-hybridization
- dicts+langpacks/language extensions de-hybridization
- design proposal for front page basic nav
- review Wil's patch for bug 400986 and come up with approach for caching problem
- search performance for what we have now in prep for Fx 3 - bug 400986
- do designs for search engines & themes browser page
- Cycle 2 priorities:
- front page navigation
- detail page redesign
- update UpdateInfoURL service
- Cycle 3 priorities:
AMO 3.5 milestones
- Working one page at a time. Trying to get it done before Firefox 3
- There are 6 edit pages, couple submit pages. After 10 pages, will be pushed "live"
- Fligtar will come back with either a preview period or beta or run in parallel
AMO 3.2 Feedback
- Reviewed feedback on Basil, Madhava, Webdev blog, MozillaZine Forums - God help me - what a painful read.
- Best comment (Madhava, this one's for you): PriceGrabber and AMO - separated at birth!
Top issues are:
- App compatibility info missing (far and away the most pressing issue)
- Easy access to "new & Updated add-ons" addons list - these are two separate things
- Frustration with Smart install button/inability to download any version/beta testing scenario/downloading incompat add-ons
- Issues with the hybrid pages (esp. themes+ext, search engines+search tools)
- Drop in theme downloads significantly (60-90%) due to theme/appearance hybrid page
- Advanced search (including app max/min version range, platform, type, etc...)
- Other often mentioned issues
- Dealing with the use case of "Should I upgrade to Fx 3 - are my favorite add-ons compatible?"
- Include date of last update info on add-ons details (x8)
- Discussions - bring them back
- Sandbox install experience & issues
- Search button since now you have to use the keyboard to perform a search
- Can't access media/common plugins
- Review count inconsistency
- Search box is too big
- Too much "white noise" graphics & fluff
- Needs to look like this
- You need to login to install stuff (since we surfaced experimental)
- I've posted to the very heated debate on MozillaZine, there are some reasoned replies coming thru now.
- AMO v3.4 will be the "Firefox 3 readiness release"
- Working on list of potential changes based on AMO 3.2 feedback
- Super drafty PRD
- Will use a release process with 2-3 week milestones
- Is Developer Tools revamp to be included in 3.4? If so, need to bump other dev/editor/admin/stats 3.4 bugs to later release. status
- API bugs have been triaged
- Need to set Target Milestone= for selected set that needs to be fixed before Fx 3 ships
- Basil is working with the Metrics/IT team to modify the Mozilla Corp stats dashboard to include some basic AMO stats info (Downloads & Update Pings) filterable by various views.
- After extensive discussions with Mitchell, we've decided to postpone implementing Google Analytics for AMO. She's published a policy on her blog. We'll roll out Omniture first and learn from that. Too many unclear issues with Google's data policy made it a stop-ship. Cost prevents us from using Omniture on AMO in the short term. IT/Ops is on board for maintaining Urchin thru 2008.
April 2nd (Post-Mortem)
- AMO 3.2.1 release scheduled for April 9th. Bugzilla Query. (You can subscribe to the AMO3.2.1 query from your bugzilla account - "Preferences", "Saved Searches" tab)
Agenda will be primarily a discussion around what worked well and what didn't in releasing AMO 3.2
- Release Definition/PRD
- "Reskin" is not the right term for the project. Should have called it "4.0". We are changing a lot of stuff. Implication of a major version change.
- Was the PRD detailed enough? Yes.
- It was mostly about redefining the page rather than what users are supposed to be doing on the page.
- Link to bugs worked well.
- We knew what the end goal was but there was implementation details that got in the way. E.g. How will search plugins work? How will hybrid pages work?
- Should have flagged the hybrid pages
- Initially started as a reskin and then migrated with the PRD (and it changed the scope).
- Install buttons was new scope but we decided that we wanted to put it in because it's the right thing.
- Mockups/UI Design
- Ones that were done were good - there was subpages (exception). Flows were good and understandable. From mocks to styles pretty easily
- Would have gotten Craig setup on AMO so that we didn't have to translate his stuff
- Try to allow old and new CSS to co-exist. We broke a lot
- I think there was some confusion of where to see the latest mocks
- Include a reference site in a website directory instead of a wiki
- Wireframes were too fleshed out. Perhaps hand off more to Henry/graphics designer
- Can speed up the process
- Craig didn't know that there is another section (dev cp) that he's not touching. Fligtar had to undo all of that for the dev pages. Having Craig be aware of what's out there already.
- Scope of Release
- Do one major page at a time. Release each page separately. Requires each template simultaneously.
- Having the pages assigned to morgamic for too long. Delegate sooner.
- Didn't pay enough attention to bug lists. Use BZ better and keep better track of stuff every week.
- CakePHP / we upgraded versions out of necessity
- PHP5 / couldn't avoid it, PHP4 is losing support
- Conflict with PHP5 & CakePHP we should have seen earlier. PHP5 should have migrated earlier or later but not simultaneously.
- Resurrect the webdev radar so that we can see all the projects and dates (migrate back from the wiki page)
- In the original plan, we should have included stuff for localizers and load testing. We needed to add those to the plan earlier in the process. (Load testing was last minute). As a webdev group, we need to figure out the load testing process prior to release. (Esp. as there's a dependency on Firefox)
- Make it smaller
- Scheduling & Estimation
- Get better at it. Morgamic was too optimistic in 2007.
- Reworked schedule was more realistic. Having smaller chunks allows to estimate better. We got Schrep/third-person feedback which was great. We learned from 2007 Q4 and got some sit down time with devlead & project manager and hash it out. Got the entire scope into it.
- Scrum/Agile might help us with smaller scoped releases and shorter durations.
- Development Processes
- Work towards a code freeze and not a release date
- Getting preview updated (we need to be diligent on what changes that break the site). All the changes that make db changes, need "try" scripts to prevent preview from getting horked.
- If anyone is blocked, it should become a priority to unblock them
- Preview: A staging server (only update and don't break it)
- Remora-reskin: What we used for instant changes (every 2 min)
- Should we automatically upgrade preview on a regular basis without breakage
- We need a daily build that everyone can look at it
- We need to find a way to version our db changes so that when we push stuff to stage that we know what structure is running on a particular site. (Need a db state + SVN rev and match them up). Drupal has install/uninstall for modules - perhaps something like that.
- Would be good to have a wiki page that has a common use case for admins & editors - monitor what they are doing most frequently (cpollett)
- Stephen felt like there's nothing happening and then way too much happening. He was verifying bugs all the way till release.
- We didn't have the test period we expected in the schedule.
- Could have had another week to test
- Time that we needed to deploy.
- Public/Beta Testing
- Preview didn't give us the audience we needed. We got only about 5-6 bugs
- We don't know how many people actually hit us
- We should put a link to the preview version from the production site
- Put a see a bug in the header/footer of the preview (or email alias) - file a bug quickly/form instead of BZ account requirement.
- Ask oremj about how upload XPI's on preview works (doesn't have production effects)
- Need to add xpi upload tests. clouserw, we have some in the remora stage test data. Please a README for those remora stage test data.
- Release Process
- oremj intimidates morgamic
- Give them a earlier headsup
- If we are going to do load testing, then, how long and when to start
- Can we use load testing daily/weekly? (req/sec - log replay, distributed load - grinder, most popularly accessed pages from stats - page load time tool)
- What worked right? / Always do this:
- Using IRC to keep constant communication is great
- Proud that we are finished with the milestone!
- What didn't work. Never do this again:
- Assuming that morgamic has a lot of dev time (bad assumption)
- We didn't do a good job of assigning stuff - critical bugs assigned to nobody@ is bad
- If we could get to the really important stuff first (install buttons) instead of polish
- Earlier is better for complex features so more eyeballs can see it
- We should communicate with the outside world (users better) - we should make AMO even more public. A lot of things changes and overwhelmed and negative comments because of amount of change. Know about the changes earlier and comment about it.
- We did a poorer job of turning user feedback into features. We can filter stuff that comes thru all channels. (e.g. http://blog.mozilla.com/webdev/2008/02/15/amo-3.2-preview/#comments)