Vote with my feet and go where? Apple Feedback reports are an infamous black hole, so not Safari. Firefox has had some bugs stay open for 15 years or more. And everything else is reskinned Chrome.
The author even linked to the jwz CADT article about how this attitude is endemic to open source (and nothing has improved in the >20 years since he posted it). There’s nowhere to go that’s any better.
Great argument to pay $ for open source software. Both parties are incentivized to submit and respond to high quality bug reports.
The Google open source model is an unfortunate one. The rest of the world is forced to compete with "funded by ad money" or to give their work away for free (Firefox, Ladybird, etc.).
Imagine anyone who uses a web browser paid $1 for a yearly license? $10B/year to fund a software company that develops it? The dream.
> Imagine anyone who uses a web browser paid $1 for a yearly license? $10B/year to fund a software company that develops it? The dream.
Mozilla got $680M in revenue in 2024 (not from $1/year licenses, but that does not matter). Do you think they would produce a 1.5x better browser if their revenue went to $10B/year?
This works with commercial software one pays for. This doesn’t work with open source software, because a user of open source software is not a customer.
I actually stopped providing free feedback to corporations unless I have a direct connection to the product team in a VIP capacity (with freebies). Otherwise, if they want my feedback, they'll have to pay me. If they don't like that, they can go search Reddit.
> it came up as a "broken image" icon (which isn't a bug) and failed to display the alt text from the img tag (which is)
The test page he provided http://perl.plover.com/NPC/test.html renders correctly for me in Chrome on Mac (Firefox renders the alt text, but no broken image icons), maybe the bug report _is_ obsolete?
I'm confused. Is this something that just happened or was it 2012 and the story was just posted yesterday for some reason?
> [ I wrote this in August 2012 and then forgot about it. ]
Edit: I think it is old based on the addenda
> A Mastodon colleague was complaining today about a similar practice, and I wanted to refer him to this article, but couldn't because I had completely forgotten to publish it.
Fair enough I guess, but odd choice to start the article with "18 months ago" and date it 18/11/2026.
...this really would be the perfect use case for AI, wouldn't it?
Of course actual code changes to something like Chromium need to be driven by humans, for all sorts of reasons. And I'm not suggesting Google should hand over all bug triage to a machine, or anything like that.
But if you've already decided, for better or worse, that you're going to auto-close old issues—maybe at least have an AI look at them first? The bug was clear and straightforward; an agent should be able to confirm that it still exists, either by reproducing it or by looking at code changes since the problem was reported. It doesn't even need to be a particularly strong model.
Surely Google of all companies has the compute for that?
Who's downvoting this? People who hate AI or people who don't mind huge backlogs of high-effort human submitted bug reports that rot for all of eternity?
It costs very little to have an AI do a first pass on bugs. Is it a dupe of another bug? Was it fixed already? Can it be reproduced? What's the estimated impact and scope of the fix? Maybe even fix it autonomously if it's low-risk. These are all things humans don't enjoy doing.
> ESR does precisely this; he uses AI to answer [...] bug reports.
I'm not even suggesting this, to be clear. If the bug report was written by a human, I think it's pretty rude to have a robot respond. I just think the robot should take a second look before something is auto-closed for being too old.
Now, if the bug report was itself written by a robot, that's another story. :)
I work at a place where if you have too many "open tickets", you get a message from someone in a managerial position telling you that you have too many open tickets ("thanks, I did not know that!"). The kicker is, however, our ticket assignment works on a best-fit basis, and I dip my hands in enough stuff a lot of cases are assigned to me, and I don't mind. What I do mind though, is having to close tickets just to close tickets to fill someone else's absurd and abstract and useless bottom-line. But the managers don't seem to understand the difference between closing tickets and helping people. I don't close tickets before the case is resolved, and I don't care that there is a ticket assigned to me five years ago about a GPU virtualisation problem that may have gone stale 2 years back when NVIDIA, bless their heart, updated their drivers. I can't attend to the ticket on a whim, and it's on a long back-log, but I _know_ the problem may still be there which is why I keep the ticket assigned to myself. And so on.
I am pretty sure Google has the same problem as I attribute to our work place (just to be clear, no I do not have a problem with too many tickets, we're storing garbage 100x the volume these tickets require as upkeep, and in a year it will be 100x more for the former).
I think it’s partially a philosophic question of “what is the purpose of a bug / ticket tracker?”
Some people treat their ticket list as a historic record of defects found, whether or not they are being actively worked on or whether or not there is even a plan to fix them. To these people, the bug list is documentation of all the software’s issues.
Other people treat the bug list as their work checklist. Tickets are a working set of things we are actively working on and plan to resolve. If you don’t have a plausible plan to fix the bug, you close it as “won’t fix”.
I find camp 1 just accumulates and accumulates and the bug list becomes this long menu that you have to periodically groom and pick from. I happen to fall in camp 2 and am very aggressive about closing things that will realistically never be fixed. My bug list is what I actually have a plan to resolve.
When you get both camps in one company, it gets confusing and you end up with these “bankruptcies.”
if it is an actual issue that ought to be fixed but you don't have the resources to do it, then it could be moved to a backlog instead closing it as "won't fix". the backlog is the documentation of all the issues and it is clear that there is no plan to fix something in the backlog. that way it works for both types.
This. From the article: "but there is no cost to keeping them around." This is true from a holistic POV of a software project as a whole. This is /not/ true for an individual in an organisation who wishes to remain employed and whose performance review depends on arbitrary metrics.
A few years back, when $CORP employed management consultants to streamline their operations, I discovered my red line: I refused point blank to lie in Jira to make graphs go in the direction the consultants needed them to instead of the way that painted an accurate picture of reality. I kicked up a huge fuss and quit over this. No-one else did - not because they agreed with the consultants, but because I was the only person on the team who was in a stable enough position to survive walking away.
I do not blame people for wanting to live in a house and maybe occasionally eat.
are these issues only you can fix? i agree that any single developer should not have to many issues assigned to them. but it depends what assigned means/ to me it means that only the assigned person and noone else is going to work on this. consequently at one company where i worked, issues did not get assigned until the person started to actively work on them. everything else was in a general pool that no one in particular was responsible for, at best with tags of who or which team this might be relevant to.
I actually turned off bug reports for my open source project, after some loser got abusive with me because I told him “that’s not a bug report, that’s a support request, and I won’t support you”.
The problem with bug reports is that the majority of them are feature requests. Or they are support requests. Very rarely are they actual bug reports.
So, I turned off bug reports.
The way I handled them before the age of bug tickets was to have a mailing list, and to tell people who emailed me privately to instead send an email to the list if they wanted support. This way, the community as a whole could either support the person, or we could have a discussion about why I wasn’t going to implement their favorite pet feature “for fun and for free”.
Right now, in the age of AI-automated attacks based on 0-days, developers need to pause adding new features and make anything Internet-facing as secure as possible (I’m very lucky my little open source project has never had a remote code execution bug).
A bug report sort of is a kind of support request, if you think about it. There's just a difference between people saying that something you intended doesn't work how you intended it, and people saying that how you intended it is not how they wanted it.
A support request is something that belonged on Stackoverflow, a “how do I...” question (and yes, I did at times respond to those kinds of “bug reports” by improving the documentation if it wasn’t already well documented).
A bug report is when something doesn’t work as specified in the documentation.
A feature request is when someone wanted my program to do something it can’t do.
This is all true, though I'd add that as I wrote it, a bug report would be asking you to support the software with a patch and a feature request would be asking you to support new functionality with an update.
We definitely did, but I think it was only for some components (tags? I forget what the monorail name was)
Auto-closing is also completely warranted imo, especially in such a large product with such wide use. Features changes, bugs morph, crashes move, people report duplicates many times over.
Relatedly, Evan did a test bankruptcy at some point (blindly delete all flaky tests) which absolutely made sense when he did it. But that idea still makes it made me feel more wicked and unholy.
Thanks for digging this up and for your helpful response! I had forgotten about that. Maybe your reply is the reason I didn't publish the original article.
I feel like this was posted disingenuously to goad the assumption it's "AI files too many bugs so Google..." but this happened 15 years ago and is not relevant to anything. There's no reason to share this on HN.
Also every large tech company I've ever worked at has done this periodically. Especially when I worked on small teams that built internal tools. We'd have years of bugs and feature requests that we'd never be able to address. Half the open tickets were from people that had left the company. So every few years management would mass-close things with a "If this is still relevant re-open it."
I tested systems for years. Declaring bugs not fixed but making them features to work around is an old practice that Microsoft and other large companies are famous for.
I've been on internal teams with internal bug trackers that have done the same thing. It's unbelievably short sighted. Same with closing a bug as "wont fix". The bug still exists, why close the ticket? Who does thst benefit and how?
In an ideal world, all bugs would be fixed, but we all know it's not the case, and at least that makes it clear. And for the team it makes backlog management easier.
What's the alternative? You can file it as "trivial" priority and exclude those from your backlog, but what have you gained?
And it makes particularly sense for internal trackers. At least with open source they're always the small chance someone decides to contribute a fix. In a company setting, no one else is going to fix your bugs. At best they will tell their manager to tell your manager to fix it. It's not like tickets cannot be reopened.
bredren | 4 hours ago
A genuine bug report is a professional contribution. These are commercial enterprises.
They should be paying people for reports in appreciation at least.
Vote with your feet.
Analemma_ | 4 hours ago
The author even linked to the jwz CADT article about how this attitude is endemic to open source (and nothing has improved in the >20 years since he posted it). There’s nowhere to go that’s any better.
levkk | 4 hours ago
The Google open source model is an unfortunate one. The rest of the world is forced to compete with "funded by ad money" or to give their work away for free (Firefox, Ladybird, etc.).
Imagine anyone who uses a web browser paid $1 for a yearly license? $10B/year to fund a software company that develops it? The dream.
warkdarrior | 3 hours ago
Mozilla got $680M in revenue in 2024 (not from $1/year licenses, but that does not matter). Do you think they would produce a 1.5x better browser if their revenue went to $10B/year?
levkk | 3 hours ago
strenholme | 3 hours ago
This works with commercial software one pays for. This doesn’t work with open source software, because a user of open source software is not a customer.
glimshe | 3 hours ago
jamie_ca | 4 hours ago
The test page he provided http://perl.plover.com/NPC/test.html renders correctly for me in Chrome on Mac (Firefox renders the alt text, but no broken image icons), maybe the bug report _is_ obsolete?
warkdarrior | 3 hours ago
(Though it is possible a future AI traveled in time to 2012 and added hallucinated code to Chrome to create this bug.)
operation_moose | 4 hours ago
> [ I wrote this in August 2012 and then forgot about it. ]
Edit: I think it is old based on the addenda
> A Mastodon colleague was complaining today about a similar practice, and I wanted to refer him to this article, but couldn't because I had completely forgotten to publish it.
Fair enough I guess, but odd choice to start the article with "18 months ago" and date it 18/11/2026.
paulirish | 3 hours ago
Turns out I personally responded. (Did not expect that)
It was a bug in WebKit that was resolved a year later.
I have no memory of the bankruptcy closures 14yrs ago. He probably has a fair point. But anyway, details there for the curious.
IMO the title should probably have a "(2012)" suffix.
(I reposted this comment for visibility.. sorry)
subarctic | 3 hours ago
mjd | an hour ago
theragra | 4 hours ago
s3graham | 3 hours ago
Wowfunhappy | 4 hours ago
Of course actual code changes to something like Chromium need to be driven by humans, for all sorts of reasons. And I'm not suggesting Google should hand over all bug triage to a machine, or anything like that.
But if you've already decided, for better or worse, that you're going to auto-close old issues—maybe at least have an AI look at them first? The bug was clear and straightforward; an agent should be able to confirm that it still exists, either by reproducing it or by looking at code changes since the problem was reported. It doesn't even need to be a particularly strong model.
Surely Google of all companies has the compute for that?
qwerpy | 3 hours ago
It costs very little to have an AI do a first pass on bugs. Is it a dupe of another bug? Was it fixed already? Can it be reproduced? What's the estimated impact and scope of the fix? Maybe even fix it autonomously if it's low-risk. These are all things humans don't enjoy doing.
strenholme | 3 hours ago
In my experience, over half of AI generated (or AI assisted) security reports are not actual security issues.
Wowfunhappy | 3 hours ago
I'm not even suggesting this, to be clear. If the bug report was written by a human, I think it's pretty rude to have a robot respond. I just think the robot should take a second look before something is auto-closed for being too old.
Now, if the bug report was itself written by a robot, that's another story. :)
rikschennink | 3 hours ago
hackrmn | 3 hours ago
I am pretty sure Google has the same problem as I attribute to our work place (just to be clear, no I do not have a problem with too many tickets, we're storing garbage 100x the volume these tickets require as upkeep, and in a year it will be 100x more for the former).
ryandrake | 3 hours ago
Some people treat their ticket list as a historic record of defects found, whether or not they are being actively worked on or whether or not there is even a plan to fix them. To these people, the bug list is documentation of all the software’s issues.
Other people treat the bug list as their work checklist. Tickets are a working set of things we are actively working on and plan to resolve. If you don’t have a plausible plan to fix the bug, you close it as “won’t fix”.
I find camp 1 just accumulates and accumulates and the bug list becomes this long menu that you have to periodically groom and pick from. I happen to fall in camp 2 and am very aggressive about closing things that will realistically never be fixed. My bug list is what I actually have a plan to resolve.
When you get both camps in one company, it gets confusing and you end up with these “bankruptcies.”
em-bee | 3 hours ago
tsukikage | 3 hours ago
A few years back, when $CORP employed management consultants to streamline their operations, I discovered my red line: I refused point blank to lie in Jira to make graphs go in the direction the consultants needed them to instead of the way that painted an accurate picture of reality. I kicked up a huge fuss and quit over this. No-one else did - not because they agreed with the consultants, but because I was the only person on the team who was in a stable enough position to survive walking away.
I do not blame people for wanting to live in a house and maybe occasionally eat.
em-bee | 3 hours ago
strenholme | 3 hours ago
The problem with bug reports is that the majority of them are feature requests. Or they are support requests. Very rarely are they actual bug reports.
So, I turned off bug reports.
The way I handled them before the age of bug tickets was to have a mailing list, and to tell people who emailed me privately to instead send an email to the list if they wanted support. This way, the community as a whole could either support the person, or we could have a discussion about why I wasn’t going to implement their favorite pet feature “for fun and for free”.
Right now, in the age of AI-automated attacks based on 0-days, developers need to pause adding new features and make anything Internet-facing as secure as possible (I’m very lucky my little open source project has never had a remote code execution bug).
LoganDark | 3 hours ago
strenholme | 3 hours ago
A bug report is when something doesn’t work as specified in the documentation.
A feature request is when someone wanted my program to do something it can’t do.
LoganDark | an hour ago
paulirish | 3 hours ago
Turns out I personally responded. (Did not expect that)
It was a bug in WebKit that was resolved a year later.
I have no memory of the bankruptcy closures 14yrs ago. He probably has a fair point. But anyway, details there for the curious.
IMO the title should probably have a "(2012)" suffix.
s3graham | 3 hours ago
Auto-closing is also completely warranted imo, especially in such a large product with such wide use. Features changes, bugs morph, crashes move, people report duplicates many times over.
Relatedly, Evan did a test bankruptcy at some point (blindly delete all flaky tests) which absolutely made sense when he did it. But that idea still makes it made me feel more wicked and unholy.
mjd | an hour ago
I'll add a note to the bottom.
gavinray | 3 hours ago
Only once have I taken the time to submit a Chrome bug report.
Chrome Canary updated and then refused to launch entirely, on Windows 11. I dumped the Event Viewer and application crash logs into the bug report.
It was fairly promptly fixed, and an overall positive experience.
- https://issues.chromium.org/issues/465823612
- https://chromium-review.googlesource.com/c/chromium/src/+/72...
settsu | 3 hours ago
motbus3 | 3 hours ago
coldfloor | 3 hours ago
Also every large tech company I've ever worked at has done this periodically. Especially when I worked on small teams that built internal tools. We'd have years of bugs and feature requests that we'd never be able to address. Half the open tickets were from people that had left the company. So every few years management would mass-close things with a "If this is still relevant re-open it."
ginthian | 2 hours ago
cosmotic | 50 minutes ago
Leherenn | 29 minutes ago
In an ideal world, all bugs would be fixed, but we all know it's not the case, and at least that makes it clear. And for the team it makes backlog management easier.
What's the alternative? You can file it as "trivial" priority and exclude those from your backlog, but what have you gained?
And it makes particularly sense for internal trackers. At least with open source they're always the small chance someone decides to contribute a fix. In a company setting, no one else is going to fix your bugs. At best they will tell their manager to tell your manager to fix it. It's not like tickets cannot be reopened.