Hacker News new | ask | show | jobs
by fiala__ 2588 days ago
The problem with YouTube specifically, is that while blog posts are easy to host on pretty much anything that has a CPU, self-hosting videos at scale is essentially impossible. So while Medium only added a marginal amount of convenience for your average blogger, YouTube straight up enabled the kind of content it hosts.
3 comments

This. And even if you manage to somehow self-host your videos, YouTube is also a discovery service for these videos. Think about how many times you looked for a video on a certain topic in YouTube's search bar itself, the recommendations and the Trending tab.
>if you manager to somehow self-host your videos...

Wait, we've been self-hosting our own videos for years. It's not exactly rocket science. YouTube is convenient, but what it is doing from a technical standpoint isn't that difficult to do on your own server.

>It's not exactly rocket science. YouTube is convenient, but what it is doing from a technical standpoint isn't that difficult to do on your own server.

You've left several comments in this thread talking down to people as if they're ignorant about how to write an HTML5 <video> tag on their own web server.

I wasn't the one that downvoted your comment but for some reason, a lot of technical folks like you misunderstand Youtube and how it enables video uploaders. (A previous commenter misunderstands Youtube the same way and my previous reply to it.[0])

There is no self-hosting web server stack to serve videos that charges $0 to the content creator whether it gets zero or 1 billion views. Therefore, repeatedly recommending "self-host your videos" -- completely misses the point.

Consider a corporate giant like Microsoft. Several years ago, they used to self-host their tech videos on channel9.msdn.com.[1] Now they're hosted on Youtube.[2] Obviously, MS is not so technically inept that they don't know how to stream their own videos! They also have billions in cash to prevent "server bandwidth exceeded" errors so cost isn't the issue.

Stop and think about why Microsoft switched to Youtube instead of using their own MS Azure infrastructure. As for the other metaphor of "sharecropper" for Youtube that seems popular... is Microsoft a "sharecropper"? Why or why not?

[0] https://news.ycombinator.com/item?id=18488275

[1] clicking any of the new videos on Microsoft's front page will play embedded Youtube videos: https://channel9.msdn.com/

[2] https://www.youtube.com/user/Microsoft/search?query=build

I think you and the other poster may be talking passed each other.

He seems to be trying to emphasize the relative triviality of the technical problem from building the MVP absent leveraging "someone else's computers" perspective.

You're talking about the delta and impact created by the fact we've become so dependent on someone else's computer to point us in the right direction to create the visibility and discoverability we want.

The point I think both of you are dancing around, but not hauling out into the light, is these platforms enable the abstraction of the techie work, and allow creators to just publish. Creators are so dependent on not having to do the techie work, that unfortunately, they are left at the mercy of the techie+business platform provider, and as a result, are vulnerable to censorship based on that platform's visibility to the world at large.

Large integrated platforms are cool and all, but at some point, we need to sit down and look at the federatability of these types of communication platforms.

We can't rely on implicit Gatekeepers not being manipulated into acting as amplifiers/dampers as circumstances warrant from their side.

At least that's the vibe I'm getting.

>The point I think both of you are dancing around, but not hauling out into the light, is these platforms enable the abstraction of the techie work,

No, not the techie work. In both of my previous comments, I de-emphasized the technical reasons.

Instead, I've tried to emphasize that the killer feature of Youtube for the content creator is the simplifying of finances down to $0 costs for distributing video. Others may argue that "audience & discoverability" is equal to (or more important than) the $0 costs to distribute. That's valid as well.

Since a technical solution of self-hosted video web stack does not solve $0 distribution costs and audience reach, it is irrelevant to the discussion. (Context of discussion was parent comments by fiala__ & ralphstodomingo talking about "scale" and "discoverability".[0])

To add some counterbalance, it does not mean Youtube's "value proposition" of $0 payment for audience reach is always a good deal. An example of this is Netflix. They don't need nor want Youtube's servers to host videos.

>Creators are so dependent on not having to do the techie work,

Again, this type of statement is evidence of techies misunderstanding Youtube.

Even if the content creator hired a techie such as a webmaster to set up a self-hosted video site, it still does not solve the problem that Youtube solves.

Even if you gave a set-&-forget "video hosting web appliance" to a content creator, it still doesn't solve the same problems that Youtube solves.

In both cases of those technical solutions, you've created new problems that the content creator doesn't want to deal with!

[0] https://news.ycombinator.com/item?id=19937280

Do you have a trustworthy source for MS "switching" to Youtube? Their self hosted videos of Build 2019 [0] is what they directed me to when i was searching for sessions. I later found Youtube versions through some other source as well (probably a link on HN), but this seems more of a PR move ("we don't do evil walled garden anymore") to me than a technical necessity.

I do agree with your core message, that video is hard and not something a small player should build their own solution for. But saying video is to hard even for Microsoft seems a bit of a stretch...

[0] https://www.microsoft.com/en-us/build eg https://mybuild.techcommunity.microsoft.com/sessions/77571

>But saying video is to hard even for Microsoft seems a bit of a stretch.

I actually said the opposite of that. I emphasized that MS had both the technical skill and money to host their own videos and yet they still moved channel9 videos to Youtube. I wanted readers to pause and think about why they did that.

By using Microsoft as an example, I was hoping to break the mental loop of always referring back to "technical issues" as the reason creators choosing Youtube. It's not technical.

https://www.YouTube.com/MicrosoftDeveloper has all the Build 2019 videos (and 2018 as well.

We still self-host for countries where YouTube isn’t available (or if something goes down), but for Channel 9 anyway[1], we’ve moved to YouTube because that’s where the audience is and it makes sense to be where our users expect us to be.

This isn’t about video being hard; our internal player is pretty good and works everywhere (which is not true for YouTube). This was a conscientious decision to go to where our audience is. We still have a self-hosted backup for each video, complete with captions in a variety of languages. But we’d heard repeatedly that our users preferred using YouTube to discover content and not being there didn’t make sense. This isn’t true for all Microsoft content, but for our developer focused videos, we want to be where the community is.

[1]: https://channel9.msdn.com

(I was directly involved in the decision to move our developer content to YouTube.)

If it was a PR move, is "we don't do evil walled garden anymore" more important than "Azure rocks and we do a lot of shit with it, look at this case study"?
Okay, you're clearly missing the point here. Sure, you can upload them videos and put them in a page somewhere, hosted by a web server somewhere. But why would you do that?

The cost of keeping your videos accessible are higher than keeping a typical blog accessible. Storage and bandwidth are obvious ones. How do you get ad networks to monetize your videos, if you want to? How'd you get people to discover your videos, as they are not first-class citizens of a search result page? Where do people even watch videos these days? Everything is on Facebook or YouTube.

You can somehow self-host your videos, but even if you manage to do so, your original purpose for doing it in the first place won't be met, it simply isn't worth it.

Contrast this with Medium, where there are not as many benefits for letting them benefit from your content. Hosting plaintext or markup is trivial, easier to maintain a blog with stuff like Wordpress than to learn how to manage your own server directly to optimize for video. There are a lot of aggregators like HN and Reddit that are populated mostly by blog content, and even if you don't resort to these, your pages are first-class eligible citizens in search results.

Heck, even the policy regarding discussions not being actively shown may hurt your specific type of content. Moderation is tolerable with Disqus, for example.

Yeah, it’s almost trivial as long as none of those videos get popular...
How do you do this? I’ve looked at Digital Ocean and their super cheap bandwidth, but storing a serious amount of data would get expensive.
Well, so is Medium. I can self host a blog but how will people find me? As the OP said, Medium ranks well.
It ranks well, but you rarely go on Medium's search bar or feed. I'd usually be here in HN, or Reddit, or some aggregator, where your self-hosted content can be found.
Is it really Impossible to selfhost videos? My 5€ webhosting has 250GB storage and unlimited traffic and every decent CMS has video capabilities.

I think the hosting is not the problem. But the social features that YouTube provides are not easily replaced...

Self-hosting some very niche videos? Sure. Trying to make a generally popular channel? Forget it.

It's the transfer and bandwidth that are the problem. There isn't really such a thing as "unlimited transfer" (there's a hidden limit past which they'll rate-limit you, and kindly ask to stop or pay more). Moreover, once your video gets somewhat popular, you'll hit bandwidth issues. Given how popularity on the Internet seems to happen in spikes, this will likely severely limit the reach of your video.

People are also spoiled by big video services with unlimited budgets and CDNs all around the world.

That's an issue Peertube [0] is trying to solve by having viewers contribute back some bandwidth via webtorrent.

[0]: https://joinpeertube.org/

If you want your media to be readily available despite your limited bandwidth the answer is torrent. If you still want to have control on who does what, then you're out of luck.
True. Unfortunately, there's another problem with p2p: https://news.ycombinator.com/item?id=19937861.
It is quite unlikely that the "unlimited" traffic you are paying for is truly unlimited. There is always a catch with such "unlimited" offers.
Also consider their automatic video encoding / re-encoding / resizing; I think there's services (e.g. Amazon Elastic Transcoder) that can do that for you, but that would require some development time of your own.

Is there a single video format supported by all browsers yet?

Those things are easily done with tools like ffmpeg and mbox. This stuff isn't rocket science. These are all solved problems and the tools are free and open-source. And yes, there are video format that are supported by all browsers.

The number of naysayers that are implying that working with videos is so hard in these comments is shocking to me for such a supposedly technical group of people.

It's better to keep in mind that promises of unlimited traffic usually countered by fine text provisions, which your hosting company will invoke once your website starts to be a nuisance for them. Various companies have different levels of tolerance, and actually it's a sort of problem with these "unlimited" offers that you won't know when it hits.
I asked them and for up to 40TB/month it's okay. This is 80000 views of 10 minutes full hd video. The chance that someone gets this popular is quite small.

If you are really really popular investing in a cdn is likely the only solution and for this you need some money, be it from advertising, donations or memberships. But this is the same on Youtube too, only that Youtube does the advertising for you and takes a cut.

Try it. Also there is Cloudflare doing a video CDN. I don't think they have a free plan with that, but it would be interesting to see what the costs actually are. There should not be a need for a CDN though if you are only showing a few videos to a few people. Video does the buffering thing so it should be possible to serve video half way across the world.

I would give it a go myself but I only have audio!

No, it's not difficult to do at all. Another commenter said "if you manage to somehow self-host your videos..." Somehow manage to self-host videos? They make it sound as difficult as managing to somehow write my own web browser from scratch. Self-hosting videos is relatively easy. The benefit of youtube isn't that it can somehow host my videos.
We now have peertube, which uses p2p to decrease the load on your server.
Unfortunately everyone (including GDPR laws) is scared of p2p
Care to elaborate? I find this rather a sweeping statement and quite inconsistent with the level of activity and progress being made in the p2p sphere.
Fair comment. Unfortunately (for me!) I'm going to have to look like a bit of a wally. I can't elaborate on it and my stance has changed in the time since I posted (while looking up my former 'facts').

1) Based on the GDPR comment. If you're really careful with how you do things the IP address could be argued to not be personal information (never let a user associate IP with any other PI - e.g. if a user can ban another from their channel something, don't let them see the underlying IP ban in place. Not sure how IPs are handled across ActivityPub).

2) People scared of p2p.. I'm still on the fence a little. For instance the opt out issue at Peertube[1] displays a snippet of p2p-fear however it should be noted that a feature to opt out of the p2p part is implemented in Peertube specifically now so may just be a non-issue. p2p fear needs to be determined with a simple "do users use this opt-out" before I can sway one way or the other

[1]: https://github.com/Chocobozzz/PeerTube/issues/685

Thanks for calling it out, I might not have found out that what I knew was outdated otherwise :)

> If you're really careful with how you do things the IP address could be argued to not be personal information

Current guidance (from the legal profession, not the HN peanut gallery) is that IPs are standalone personal data.

You got a source on that if you're going to be referring to the rest of us as a peanut gallery?

Everything I know and have found says IPs are classed as personal data only if combined with other data. I've yet to find anything that supports your claim, which is why I changed my mind on this.

In which jurisdiction? i.e. That's a hell of a generalisation.
p2p is widely useful technology, but in the minds of regular people, it's primarily associated with movie and videogame piracy. I don't know how common this is world-wide, but at least in my country it's widely known among people who use computers that downloading or streaming a copyright-protected video from a random site is legally in the clean, but p2p may land you in jail or have you paying fines - because it's unlawful distribution, not consumption, of copyrighted material that's punished.