That's not an accurate assessment of what's going on. The post linked states "no contact with the maintainers" not just inactivity, and if you follow the github issues linked, it's about people wondering what's going on and if anyone else can approve pull requests because there's lots of pull requests waiting. There's 843 pull requests at this time, and I just looked and over 50 are from just the last month.

It's not that there's repo inactivity, it seems to be that this is an extremely active repo which saw everything grind to a halt when the admins went dark. That's quite a bit different then just "inactivity".

> There's 843 pull requests at this time, and I just looked and over 50 are from just the last month

That's kinda overwhelming though ... imagine that if the maintainer pops up somewhere, suddenly 100 motivated people may chime in "hey please review this important pull request that's been sitting over here for a while".

There are some kinds of open source projects that are prone to this ... some are really not so bad to maintain if you have the right kind of discipline, because they converge on a stable set of functionality and platform compatibility evolves slowly, but some just naturally have endless room for variations and special cases, and as users increase, PRs increase linearly (instead of sub-linearly as you'd hope). I'm thinking in particular of https://github.com/oauth2-proxy/oauth2-proxy (of which I contributed to an older fork)

youtube-dl relies on up-to-date definitions of sites in order to download the content.

When a webpage changes layout, youtube-dl needs updated as well.

We're talking mostly about a list of site definitions more than we are core development.

Months before the lawsuit, youtube-dl's maintainers frequently closed issues that reported ongoing breakage without giving a reason. Here is one especially illuminating example:

https://github.com/ytdl-org/youtube-dl/issues/23860

And there is also an entire fork that fixes the support for just a single provider, NicoNico, because the maintainers ignored its issues.

https://github.com/animelover1984/youtube-dl

A quote from its README:

All code in this project is licensed solely with the condition that any portion of it is not permitted to be used in the main youtube-dl fork, either directly or indirectly. It is also not permitted to be used in any project that contains contributions from either remitamine or dstftw.

The two users mentioned are or were previously major contributors to youtube-dl.

It seems that youtube-dl was already a dysfunctionally managed project at the time of the lawsuit and happened to ride out on the good PR for a couple of months, before returning to stagnation once again.

To me it sounds like a plugin system would have prevented centralization and the need for forks, but would have made distribution harder for average users.

Indeed, this has been years in the making. Maintainer activity has been slowly dwindling while would-be contributors were driven away by the maintainers’ lack of communication and abrasiveness. I myself have had pull requests languishing there for years with nobody bothering to review them. Other people had their issues closed with no explanation. It was just a matter of time. Good thing that the forks have sprung up some time before upstream development halted entirely.

Then the repo might be dead because of that. I don't think people are (or should be) necessarily making assessments as to the reasoning of the admins without evidence, and there's a bunch of reasons why they might not be able to participate for a few months or even ever again, but if they are unavailable for a few months and nobody can even explain why, the end result is functionally the same if they're the only ones with the keys to the kingdom, which is that it's dead until something changes.

Is it a way to work on FOSS in decentralized or even distributed manner? Without relaying on the Github index as the critical point.

Sure, fork it and have alternate implementations and releases. This will necessarily fragment the namespace for it though, and the only way around that is to have some shared responsibility for making releases, and at that point, you might as well have enough shared admin responsibilities on the one repo, and the problem is also solved.

Another way to look at it is that this is not a problem of lack of solutions, it a problem that the person or people at the bottleneck (merging pull requests, testing and releasing) haven't shared that responsibility with enough additional people so this isn't a problem.

That, or there's some legal case in the background and they are under advisement to not work on or respond to anything to do with this repository. That seems just as likely in my eyes. Doesn't change the situation though, this repo is effectively dead right now (but someone could fork and start applying an releasing if they wanted AFAICT).

2 months is a long time in youtube-dl world. It's not really a "software project" in the traditional sense, where you can stop working on it once it's "done". It's more of a "social project", a focal point for the required ongoing activity necessary to keep sites working. Youtube-dl without daily commits is useless.

I'm currently using it to download a youtube video. If it still works for it's main function, maybe it's just not a high priority project until something breaks?

It's actually already broken. If you try to download more than 3-16 videos (the limit is not clear), you start to be rate limited to 300 kbps or so. According to Reddit this is fixed in a fork called yt-dlp

To be honest that sounds fine. You're presumably downloading it to watch it offline, the download speed isn't really material as long as it finishes eventually

youtube-dl is also used by video players like mpv for live streaming. In this case, 70 kbps is completely unusable for the vast majority of videos.

One other workaround is to use '-F' to get a list of formats, and choose one of the non-default formats to download. This seems to work for me.

Primary function, perhaps, but it's used for lots of websites that aren't YouTube, and some of those websites have broken.

There are open pull requests and/or bugs for many websites that aren't being approved. This is rather unusual for youtube-dl, it normally had a release ever two weeks or so.

It may say “YouTube” in the name, but it’s a tool designed to work with many sites.

For a project like youtube-dl it is a long time, because they use unofficial APIs (fancy word for scraping) of video sites that can shift even on daily basis. If you look at their Github issues it is just people endlessly complaining that some websites are broken again

> unofficial APIs (fancy word for scraping)

Using a non-public API is not at all the same as scraping, which refers to parsing a rendered HTML page for the content you want.

Both have this maintenance problem, but one's not a fancy word for the other.

That used to be true, but today, with so many websites operating as SPAs against undocumented APIs, I think it's reasonable to redefine "scraping" to mean extracting data from unofficial APIs in addition to extracting it by parsing HTML.

After all, what is a scrapeable HTML page if not a grotesquely convoluted undocumented API with an unstable output format?

Scraping refers specifically to extracting data from a format designed to be read by humans instead of machines.

The gross inefficiency and low data-to-layout ratio are the key things being expressed through connotations of the word "scrape". To scrape is to extract a small amount of something from a much larger substrate.

To call every query a scrape is to diminish the specificity and utility of the term.

If an unofficial API returns JSON that looks like this:

    {
      "id": 3422,
      "title": "My essay about cheese",
      "published": "13th August 2021 at 3:45pm",
      "abstract": "<p>In which I write about cheese!</p>"
    }

And I write code against that which includes stripping the HTML tags from "abstract" and converting the date format in "published" into in ISO datetime... am I writing scraping code?

I would argue that I am, even though it started out as a JSON wrapper.

"To call every query a scrape is to diminish the specificity and utility of the term."

Absolutely disagree with you there. I interpret the term "scraping" as "writing code that gathers data from a source that has not deliberately published that data in a usable format". Gathering data from any kind of API fits that criteria for me, since most APIs only give you a subset of the data at a time.

I think the reason I care so much about this is that I coined the term "git scraping" to cover a variant of scraping that uses Git repositories to store the data and track changes over time - and git scraping applies equally to data sourced from APIs as it does to data sourced from HTML pages. https://simonwillison.net/2020/Oct/9/git-scraping/

If everyone insists that is what it means for long enough, then that is what it will mean.

The term was coined to differentiate how difficult it is to extract data from a format that was patently not intended to efficiently spread raw data to other machines. If that meaning erodes, and it's just yet another way to say an API query, it will be a great loss for the precision of our terminology.

Disagree. I see scraping as about obtaining data in bulk that hasn't been deliberately packaged up for you to use as-is.

Most APIs are not designed to give you all of the data at once - they exist to serve other purposes, usually involving returning a small subset of the data to power a user-facing feature.

If someone asks me "where did you get those Olympic medal results?" and I say "I scraped them" I think that's accurate vocabulary whether I parsed HTML or gathered them from hundreds of undocumented API calls.

If I had downloaded a neat CSV file from the Olympics website with all of the data I needed in one go I wouldn't feel comfortable calling it scraping.

Re-reading your comment, I think what I'm describing here does actually fit with your "how difficult it is to extract data from a format that was patently not intended to efficiently spread raw data to other machines" definition - except I'm including APIs that return only a subset of the data as part of those inefficiencies in obtaining the raw data.

APIs are rate-limited all the time. Compensating for being throttled or paginated isn't scraping.

Scraping is the act of extricating data from the layout and markup metadata meant to make it pretty for humans.

APIs generally don't include any of that, your HTML-in-a-JSON-object example notwithstanding.

I'd have no objection to calling it scraping when you strip those <P> tags, but aggregating the results of several API queries is bog-standard textbook API usage, which we use the term scraping to differentiate from.

I think we can at least agree that there is no formal definition of "scraping".

That said, I had a look around and the definitions I could find tended to support my interpretation:

https://en.wikipedia.org/wiki/Web_scraping - "Newer forms of web scraping involve monitoring data feeds from web servers. For example, JSON is commonly used as a transport storage mechanism between the client and the web server."

https://towardsdatascience.com/web-scraping-basics-82f8b5acd... - "There are 2 different approaches for web scraping depending on how does website structure their contents." (HTML scraping and API access)

https://realpython.com/beautiful-soup-web-scraper-python/ - "Web scraping is the process of gathering information from the Internet. Even copying and pasting the lyrics of your favorite song is a form of web scraping! However, the words “web scraping” usually refer to a process that involves automation"

The more formal dictionaries (Merriam Webster and suchlike) don't seem to have formed an opinion on this one yet!

I think the definitions you're cherry-picking are examples of the erosion of the term's specificity, We need a word to describe scraping data values from a body of human-oriented markup. These folks are pushing the limits of ambiguity to rob us of that, and our reward is yet another word that just means using any API.

What is the upside of using this word in such an oddly vague, expansive way? What happens to those of us who need to convey the original specific meaning we coined it for in the first place?

"We need a word to describe scraping data values from a body of human-oriented markup"

I call that parsing, which is one of the steps in scraping that may or not be necessary depending on the data source.

I need a term that means "using automation to gather data from the web, when that data has not been published in a way that is suitable for my purposes". Scraping works great for that!

So you get a video feed in the end, that is viewed by human presumably?

What you do with the data after you have it is not a qualification for whether or not you got it by scraping.

Sounds like an exhausting thing to maintain. It's not like writing scraping (or even just changing slight variations in an API) is terribly interesting.

True, but it’s also the kind of product that’s instantly useful and itch-scratchy. Youtube-dl not working on the video you’re downloading today? Well if you’re a maintainer you can just patch it yourself. (Non-maintainers can too of course, but I imagine the maintainers have the know-how to actually fix things)

The amount youtube and the other sites it supports changes to subvert the use of this very tool, I can see why people are wondering what's happening

Having contributed to youtube-dl in the past, long turnaround times from the maintainers was pretty normal. I've had (and still have some) open PRs that have been ready to merge for going on a year. The two months is really not that big of deal.

That being said, the project probably could use some reorganization. It requires a lot of community contributions to keep all the extractors maintained so long turnaround times for reviews isn't ideal.

> Considering there's been weekly releases for years, two months is odd and worry some.

https://old.reddit.com/r/DataHoarder/comments/p9riey/youtube...

Sounds more like maybe the person is sick or something.

As other mentioned, it’s different with YouTube, as the site changes constantly

Also, in common with other massive online properties for instance Amazon, not consistently: changes sometimes roll out a bit at a time so users in different areas get different versions, either due to global roll-out being a staged process or because a UI experiment is being performed. It usually doesn't matter to a well written scraper as the core data is still accessible in the same way despite the UI sugar coat having changed, but sometimes there are significant enough changes under the hood too that the scraper needs to deal with while still supporting the older format(s).

Exactly. It is ok to compete but announcing your competitor project is dead to promote your own, it is a little sketchy to say the least.

I guess it depends on if sub-2-month feature development is needed to keep up with YouTube's changes or not.

I maintain a GCC code coverage tool on GitHub, and since GCC doesn't change very often and the feature set of the tool is fairly complete, I sometimes go 6+ months without commits. Usually I don't touch it unless someone opens an issue.

I mean, just look at <https://github.com/ytdl-org/youtube-dl/issues/29809>.

They don’t even bother removing spam.

Seriously. I've seen core/lead maintainers on projects I worked on go awol for months and years at a time, and come back guns blazing.

dark thought but I've had multiple instances in last year+ where someone hasn't posted in awhile, tweeted etc and I wonder for a second, maybe they died? The pandemic is real. And random ppl disappearing unexpectedly is part of it. I hope all is well.

It's August. Could the maintainers be on holiday?