Intent to de-support: traditional add-ons
sancus at thunderbird.net
Fri Nov 1 21:56:52 UTC 2019
Sorry it's taken me a month to reply to this post. I don't always have
time to dedicate to statistics work and these sorts of lengthy
discussions. It took me roughly 2.5 hours to dig through numbers and
server logs and write this email.
On 10/3/2019 12:35 PM, Eyal Rozenberg wrote:
> Andrei, with due respect, I'd like to partially disagree with what you
> wrote, particularly (but not only) regarding statistics.
> Before doing so, however, I want to ask the list of some clarification
> regarding the methodology with which we gather usage statistics: Under
> what circumstances; which users have stats collected; what happens to
> people whose older versions don't know about ATN; etc. If someone could
> write a few paragraphs about this (or send a link) I'd be grateful.
All versions of Thunderbird ping ATN, because requests to the old AMO
domains are redirected to us. We even get pings from a few stragglers on
versions in the single digits.
You can find global Thunderbird statistics on stats.thunderbird.net - in
particular, the percentage of users on each version
<https://stats.thunderbird.net/#version>. There are roughly 1.5% of
users on versions 38 and below. 2.5% on 45.*, and 3.5% on 52.*. Plus
maybe 0.5% on random older betas, mostly 56 because of a weird auto
update rule. In total, 8% on "old versions"(prior to 60). While quite
low overall, It's higher than we would like, but a bit lower(~6.5%) on
weekends, probably due to users stuck on employer-installed versions.
> 1. Quoting an annual 1-2% userbase growth rate does not say much about
> what happened over the past few months, when a lot of extension support
> was lost. Do you have stats for that?
Nothing has happened over the last few months, as you can see
<https://stats.thunderbird.net/>. Extension support is not significant
to most users. Seasonal variations like summer European vacations,
Christmas, and Easter have more effect on Thunderbird user numbers than
any change to extension support ever could. Users did not significantly
change during last's year's TB60 addonpocalypse either, so I wouldn't
expect any huge difference when 68 auto updates go out.
> What that means is that treating relatively-low-usage extensions as
> non-representative is myopic; and not significantly catering to
> extensibility blocks most of the potential of user base increase.
Sure, I agree with you. The context here is not "should we support
extensions or not?" it's - how much work do we put into supporting them.
We already put in a lot of work, arguably more than the return on
investment. Add-on developers aren't poorly supported. Add-ons occupy a
massive amount of work and discussion time from Thunderbird employees.
If I had to make a wild guesstimate based on my time, Geoff's time,
Magnus' time, etc, I would say that add-ons consume somewhere around 30%
of our overall developer time. Magnus could probably give you a better
estimate. I know I spend roughly 50% of my time working on add-on
related issues and support. Regardless of the specific number, it is
quite a bit.
You interpret the substantial support that add-on developers receive as
insufficient, not realizing that it's almost more than we can afford
> 3. 1-2% annual user base growth is more of a stagnation than growth a
> slight waning of the user base; the population in countries in which TB
> is popular is growing at a higher rate than that, I would guess. I'm not
> saying that's a bad thing considering the objective circumstances,
> 4. There is a huge difference between the part of the user base who use
> Thunderbird because they made a decision to do so, and people who have
> had Thunderbird pre-installed on their machine for some reason (e.g.
> because they're in some organization with wise IT people). Making claims
> regarding whether people tend to use or not use extensions, whether we
> have growth or decline etc. should really be treating those two
> categories of users separately - even if we can't place users in these
> categories explicitly.
True, but irrelevant. Add-ons have never, and will never drive user
growth. They may help in small ways with user retention.
> 6. "the 5-10% of users who even install and use add-ons that aren't
> Lightning." - that's a bold statement, that seems to contradict even
> Christopher's partial numbers. Show your stats on this please.
In what way does it contradict them? Keep in mind that user counts for
each add-on count the same users many times, because you are counted
repeatedly for each add-on you have installed.
Even assuming the overall average is 2 add-ons per user, if you sum the
add-ons that aren't Lightning you get 3.2M daily users, divided by 2 for
multiple counts, which would be ~16%.
However, users are heavily concentrated in the top 10-20 add-ons, many
of which either should be redundant with core, or part of core in some
way(ImportExportTools, Provider, Manually Sort Folders) or are already
on their way to core(Enigmail). If you sum outside of the top 10,
there's 1.78M add-on users, again divided by 2 to account for multiple
users, and you're already below 10%. Provider and ImportExportTools
together account for 23% of all add-on users by themselves.
I do suspect that this guess overstates the number of individual users,
because I think the average is higher than 2. Calculating statistics on
number of users that have an add-on installed is possible, but it
requires setting up some new logging and querying/compiling of those
statistics. I have added it to my list of tasks, you can track when this
work is completed in this issue on GitHub
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the tb-planning