Re: [Hls-interest] HLS Content Steering v1.0b1 preliminary specification

[Roger Pantos <[e-mail]>]

There are a few options for an engaged client that do not require client-side API in AVFoundation (and, eventually, in JavaScript). Clients that are in a position to proxy the master playlist can compose their preferred version of the steering manifest into a data:// URI and supply it as the SERVER-URI attribute, possibly adding a RELOAD-URI to it so that subsequent requests go to the “real” steering server. A data:// URI will immediately win the race, causing selection of the preferred pathway with a minimum of thrash. A similar technique for an AVFoundation-specific master playlist is to have a SERVER-URI with a custom scheme, which would be supplied by the AVAssetResourceLoader delegate.

That said, if you’d like to see a specific AVFoundation API for controlling this I’d encourage you to file a bug asking for it: https://feedbackassistant.apple.com. As always, explaining how such a feature would improve the user's experience and why it’s better than the existing alternatives will help us to prioritize it.


thanks,

Roger.

> On Mar 1, 2021, at 8:49 AM, David Hassoun <[e-mail]> wrote:
> 
> This is in part why I'm still in favor of an optional client API that can apply the data done from the steering service. It wont affect VST if this client already has the data up front to adjust the initial steering and avoids another API request. 
> 
> I do think it is very important to have the option to provide up front steering control outside of the initial playlist control as realistically many solutions wont have dynamic server generated playlists (at least for some time).  I know there is pushback to have client API when everything is being primed to be done from the steering manifest - aka server side. Though I still find it totally reasonable to have the server be default with the option to override in client. There was discussion about a client API to at minimum force a steering load and then with the local proxy we could do it all client side - or at least partial client with logic to determine when to override or not. However that still does not address a clean up front steering control. I suppose we could have an API/Boolean-Property to determine if a steering manifest is required up front before playback. If so it will attempt the request and the client can proxy/intercept to set data immediately or let the http call go through. It's fairly complex but would potentially achieve the benefit and control needed?
> 
>  <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=icon>	Virus-free. www.avast.com <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=link> <x-msg://11/#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
> On Sun, Feb 28, 2021 at 6:59 AM Daniel Weinberger <[e-mail] <mailto:[e-mail]>> wrote:
> Hi Will,
> 
> I want to add some thoughts on your last part that a player should be required to resolve the steering manifest before loading any media playlist.
> 
> I'm also in favor of clarifying the expected client behavior, but in this case there is one negative side effect of your suggestion: It delays the startup time of a video. 
> The steering manifest is referenced in the master playlist and if the player is required to resolve it before the first media playlist is loaded, this is one additional blocking step in between. I'm sure all player vendors work hard to minimize startup time and adding additional blocking requests here is not desired. 
> It can be argued that this is just a small text file and won't take long to download even with bad network throughput, but there are also environments where this is considerable: environments with higher latency.
> 
> Regards,
> Daniel
> _____________________________
> 
>  <http://www.bitmovin.com/>
> 
> Daniel Weinberger
> Principal Solutions Architect
> 
> 
> On Sat, Feb 27, 2021 at 12:39 AM Law, Will <wilaw=[e-mail] <mailto:[e-mail]>> wrote:
> Another question on this new proposal: can the spec clarify whether the client is required to request the steering manifest and resolve the pathways before it loads any media playlists? Given the playlist below
> 
>  
> 
> #EXTM3U
> 
> #EXT-X-CONTENT-STEERING:SERVER-URI="https://example.com/steering?video=00001 <https://urldefense.com/v3/__https:/example.com/steering?video=00001&session=123__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44qtYzaA4$>",PATHWAY-ID="My-Great-CDN"
> 
> #EXT-X-STEERING-TEMPLATE:ID="1",PREFIX="https://a.example.com/video <https://urldefense.com/v3/__https:/a.example.com/video__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44mqwdeF4$>"
> 
> #EXT-X-STREAM-INF:BANDWIDTH=200000,STEERING="1"
> 
> https://a.example.com/video/my-movie/mid/index.m3u8 <https://urldefense.com/v3/__https:/a.example.com/video/my-movie/mid/index.m3u8__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44hforWWE$>
> #EXT-X-STREAM-INF:BANDWIDTH=400000,STEERING="1"
> 
> https://a.example.com/video/my-movie/high/index.m3u8 <https://urldefense.com/v3/__https:/a.example.com/video/my-movie/high/index.m3u8__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44mfpy4iP$>
>  
> 
> a client could start by immediately loading https://a.example.com/video/my-movie/mid/index.m3u8 <https://a.example.com/video/my-movie/mid/index.m3u8>, and then resolving the steering manifest in the background after it had built up a few seconds of playback buffer. Players would be tempted to do this as it would provide a lower VST (by one RTT and whatever processing time there is on the steering server). This would be a bad outcome for two reasons a) For QoE as there could potentially be two cold hits against CDNs in the first 10s of playback   - one for the default and the second as the new pathway kicks in  and b) if you are trying to move away from an underperforming default pathway your startup sequence would continue to be pinned against it.
> 
>  
> 
> The current version of the steering draft is ambiguous. It says “When the player needs to poll for Pathway updates,", but does not dictate when this should happen with respect to conventional start-up.
> 
>  
> 
> Suggestion is that if the player understands steering, it MUST resolve the steering manifest prior to loading any media playlist.
> 
>  
> 
> -Will
> 
>  
> 
>  
> 
>  
> 
>  
> 
>  
> 
>  
> 
>  
> 
>  
> 
>  
> 
>  
> 
>  
> 
>  
> 
> From: Roger Pantos <rpantos=[e-mail] <mailto:[e-mail]>>
> Date: Friday, February 26, 2021 at 9:42 AM
> To: "[e-mail] <mailto:[e-mail]>" <[e-mail] <mailto:[e-mail]>>
> Cc: http-live-streaming-review <[e-mail] <mailto:[e-mail]>>
> Subject: Re: [Hls-interest] HLS Content Steering v1.0b1 preliminary specification
> 
>  
> 
> One objection we’ve heard to the approach below is that it requires the master playlist to be customized on a per-pathway basis. To address that we’re tweaking the proposal slightly: we're making the PATHWAY-ID attribute optional. If it is not present in the master playlist, the client will omit the _HLS_pathway query parameter from the steering manifest request until it obtains its first steering manifest.
> 
>  
> 
> (There are other ways for providers to communicate the initial pathway to the steering server. For instance, the SERVER-URI could be relative to the master playlist; each pathway could HTTP-redirect it to the actual steering server along with an origin identifier.)
> 
>  
> 
> Here are two examples of the current approach. The first one uses absolute URLs in the master playlist, as you might have with a dynamically-generated master playlist that knows the current pathway when the playlist is produced. The second uses relative URLs, as you might have for a static master playlist that is the same on all pathways.
> 
>  
> 
> In both cases the content is available here:
> 
>  
> 
> On the first CDN at -
> 
>  
> 
> https://a.example.com/video/my-movie/master.m3u8 <https://urldefense.com/v3/__https:/a.example.com/video/my-movie/master.m3u8__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44qV1Gx_j$>
>  
> 
> and on the second CDN at -
> 
>  
> 
> https://b.example.com/backup-video/my-movie/master.m3u8 <https://urldefense.com/v3/__https:/b.example.com/backup-video/my-movie/master.m3u8__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44llkyj51$>
>  
> 
> -----
> 
>  
> 
> Absolute URIs (master playlist URI https://a.example.com/video/my-movie/master.m3u8 <https://urldefense.com/v3/__https:/a.example.com/video/my-movie/master.m3u8__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44qV1Gx_j$>)
> 
>  
> 
> #EXTM3U
> 
> #EXT-X-CONTENT-STEERING:SERVER-URI="https://example.com/steering?video=00001 <https://urldefense.com/v3/__https:/example.com/steering?video=00001&session=123__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44qtYzaA4$>",PATHWAY-ID="My-Great-CDN"
> 
> #EXT-X-STEERING-TEMPLATE:ID="1",PREFIX="https://a.example.com/video <https://urldefense.com/v3/__https:/a.example.com/video__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44mqwdeF4$>"
> 
> #EXT-X-STREAM-INF:BANDWIDTH=200000,STEERING="1"
> 
> https://a.example.com/video/my-movie/mid/index.m3u8 <https://urldefense.com/v3/__https:/a.example.com/video/my-movie/mid/index.m3u8__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44hforWWE$>
> #EXT-X-STREAM-INF:BANDWIDTH=400000,STEERING="1"
> 
> https://a.example.com/video/my-movie/high/index.m3u8 <https://urldefense.com/v3/__https:/a.example.com/video/my-movie/high/index.m3u8__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44mfpy4iP$>
>  
> 
> The client requests 
> 
>  
> 
> https://example.com/steering?video=00001&_HLS_pathway=“My-Great-CDN <https://urldefense.com/v3/__https:/a.example.com/steering?video=00001&_HLS_pathway=**BMy-Great-CDN__;4oCc!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44h1w5BH2$>"
> 
>  
> 
> And gets back this steering manifest:
> 
>  
> 
> {
> 
>   "VERSION": 0,
> 
>   "TTL": 300,
> 
>   "RELOAD-URI": "https://example.com/steering?video=00001&session=123 <https://urldefense.com/v3/__https:/example.com/steering?video=00001&session=123__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44qtYzaA4$>",
> 
>   "PATHWAYS": [
> 
>     {
> 
>       "ID": "My-Great-CDN",
> 
>       "TEMPLATES": {
> 
>         "1" : {
> 
>           "PREFIX": "https://a.example.com/video <https://urldefense.com/v3/__https:/a.example.com/video__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44mqwdeF4$>"
> 
>         }
> 
>       }
> 
>     },
> 
>     {
> 
>       "ID": "My-Backup-CDN",
> 
>       "TEMPLATES": {
> 
>         "1" : {
> 
>           "PREFIX": "https://b.example.com/backup-video <https://urldefense.com/v3/__https:/b.example.com/backup-video__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44gKCaDMb$>"
> 
>         }
> 
>       }
> 
>     }
> 
>   ]
> 
> ]
> 
>  
> 
>  
> 
> -----
> 
>  
> 
> Relative URIs (master playlist URI https://a.example.com/video/my-movie/master.m3u8 <https://urldefense.com/v3/__https:/a.example.com/video/my-movie/master.m3u8__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44qV1Gx_j$>)
> 
>  
> 
> #EXTM3U
> 
> #EXT-X-CONTENT-STEERING:SERVER-URI="https://example.com/steering?video=00001 <https://urldefense.com/v3/__https:/example.com/steering?video=00001&session=123__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44qtYzaA4$>"
> 
> #EXT-X-STEERING-TEMPLATE:ID="1",PREFIX=""
> 
> #EXT-X-STREAM-INF:BANDWIDTH=200000,STEERING="1"
> 
> /mid/index.m3u8
> 
> #EXT-X-STREAM-INF:BANDWIDTH=400000,STEERING="1"
> 
> /high/index.m3u8
> 
>  
> 
> In this case PREFIX=“” indicates that the steerable part is the entire (relative) URI. So after the client receives the steering manifest (same as above), if it has to switch to My-Backup-CDN it will just prepend its PREFIX to the steerable part of each URL to produce:
> 
>  
> 
> https://b.example.com/backup-video <https://urldefense.com/v3/__https:/b.example.com/backup-video__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44gKCaDMb$>/mid/index.m3u8
> 
>  
> 
> https://b.example.com/backup-video <https://urldefense.com/v3/__https:/b.example.com/backup-video__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44gKCaDMb$>/high/index.m3u8
> 
>  
> 
> Unless anybody spots any other major flaws in this approach, we’ll plan to update the preliminary spec and send out another version.
> 
>  
> 
>  
> 
> Roger.
> 
> 
> 
> 
> On Feb 19, 2021, at 7:07 PM, Roger Pantos <[e-mail] <mailto:[e-mail]>> wrote:
> 
>  
> 
>  
> 
> 
> 
> 
> On Feb 17, 2021, at 1:46 PM, Pieter-Jan Speelmans <[e-mail] <mailto:[e-mail]>> wrote:
> 
>  
> 
> Hi all
> 
>  
> 
> I am pretty concerned about the backwards compatibility problems when attempting to use this in combination with older devices. It will force services to either provide two playlists, one with and one without EXT-X-DEFINE with client-side detection of the maximum version supported by the client. While this will not be an issue for up to date Apple devices, this will definitely be an issue for out of date devices but also most smart TVs, Chromecasts, Airplay implementations by those other than Apple (f.e. on LG), ...
> 
>  
> 
> Well, we’re still early enough that we could consider changing the format. I’m not particularly moved by the hardship of supporting EXT-X-DEFINE — anybody who is implementing content steering is going to have to do at least that much work in their m3u8 parsing anyway — but I acknowledge that requiring content vendors to provide a different master playlists for old, un-updated clients is a non-trivial burden.
> 
>  
> 
> What if we took the following approach instead?
> 
>  
> 
> First, add a new STEERING attribute to regular EXT-X-STREAM and MEDIA tags:
> 
> #EXT-X-STREAM-INF:BANDWIDTH=200000,STEERING=“1”
> https://a.example.com/video/mid/index.m3u8 <https://urldefense.com/v3/__https:/a.example.com/video/mid/index.m3u8__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44myYNLnQ$>
> 
> #EXT-X-STREAM-INF:BANDWIDTH=400000,STEERING=“1”
> https://a.example.com/video/high/index.m3u8 <https://urldefense.com/v3/__https:/a.example.com/video/high/index.m3u8__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44j6ekA61$>
> 
> This links the variants (and renditions) to a steering template definition:
> 
> #EXT-X-STEERING-TEMPLATE:ID=“1”,PREFIX="https://a.example.com/video <https://urldefense.com/v3/__https:/a.example.com/video__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44mqwdeF4$>”
> 
> #EXT-X-CONTENT-STEERING:SERVER-URI="/steering?video=00001",PATHWAY-ID="My-Great-CDN"
> 
> The manifest would change a bit:
> 
> {
> 
>    "VERSION": 0
> 
>    "TTL": 300,
> 
>    "RELOAD-URI": "https://example.com/steering?video=00001&session=123 <https://urldefense.com/v3/__https:/example.com/steering?video=00001&session=123__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44qtYzaA4$>",
> 
>    "PATHWAYS": [
> 
>        {
> 
> "ID": "My-Great-CDN",
> 
> "TEMPLATES": {
> 
>   "1" : { 
> 
> “PREFIX”: "https://a.example.com/video <https://urldefense.com/v3/__https:/a.example.com/video__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44mqwdeF4$>" 
> 
>   }
> 
> }
> 
>        },
> 
>        {
> 
> "ID": "My-Backup-CDN",
> 
> "TEMPLATES": {
> 
>   "1" : { 
> 
> “PREFIX”: "https://b.example.com/backup-video <https://urldefense.com/v3/__https:/b.example.com/backup-video__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44gKCaDMb$>" 
> 
>   }
> 
> }
> 
>        }
> 
>    ]
> 
> }
> 
>  
> 
> The client would load and play the master playlist as usual. When it parsed the EXT-X-STREAM-INF and EXT-X-MEDIA tags it would see the STEERING attribute and verify that there is a corresponding EXT-X-STEERING-TEMPLATE and that the PREFIX of the steering template matches the start of the URI (fatal error if either check fails). It would store away the remainder of the URI as the “steerable part.”
> 
> When it loaded the manifest it would check that each pathway had an object in its TEMPLATES object for each STEERING-TEMPLATE (matching the key to the ID). (If it did not, steering would be disabled.)
> 
>  
> 
> To switch pathways it would concatenate the PREFIX from the TEMPLATES object with the steerable part of each URI. After that, everything would behave the same way.
> 
>  
> 
> This approach isn’t as flexible as general variable substitution. But I think that it would still cover the majority of the substitution use cases (and I’d be interested to hear about those that it doesn’t).
> 
>  
> 
>  
> 
> Roger.
> 
> 
> 
> 
>  
> 
> My personal opinion is that the value of backwards compatibility would warrant the drawbacks I currently see. 
> 
>  
> 
> Best regards,
> 
> Pieter-Jan
> 
>  
> 
>  
> 
> Pieter-Jan Speelmans
> CTO
> 
> THEO Technologies NV
> w | www.theoplayer.com <https://urldefense.com/v3/__http:/www.theoplayer.com/__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44kRAt7LO$>
> Philipssite 5, Bus 1. 3001 Leuven, Belgium
> 
>  <https://urldefense.com/v3/__https:/d2oj23ymxjbl3l.cloudfront.net/__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44uAXLPbU$>
> Leuven (Belgium) - New York (US) - Singapore (SG) - Barcelona (ES)
> BTW: BE 0847.829.290, RPR: Leuven
> 
>  
> 
> On Wed, Feb 17, 2021 at 9:57 PM Roger Pantos <rpantos=[e-mail] <mailto:[e-mail]>> wrote:
> 
> Hello hls-interest,
> 
> As discussed on the call today, here is the preliminary specification for the new Content Steering feature. We are converging on the final design for 1.0, so please take a look at it when you can and tell us what you think. 
> 
> You can send feedback to this list, or to [e-mail] <mailto:[e-mail]> if you want it to go directly to us.
> 
> 
> 
> 
> regards,
> 
> Roger Pantos
> Apple Inc.-- 
> Hls-interest mailing list
> [e-mail] <mailto:[e-mail]>
> https://www.ietf.org/mailman/listinfo/hls-interest <https://urldefense.com/v3/__https:/www.ietf.org/mailman/listinfo/hls-interest__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44nsxABxl$>
>  
> 
> Legal Notice 
> 
> You can find the latest terms and policies of THEO Technologies under www.theoplayer.com/terms <https://urldefense.com/v3/__http:/www.theoplayer.com/terms__;!!GjvTz_vk!E3AorzSNM8BjAiM9Qaz2_XmVx2NyOl_J1ZzrkyQCOBZuw0Khr__44hw7Mpir$> including our "GENERAL TERMS AND CONDITIONS". In the absence of a signed agreement between you and THEO Technologies, by replying to this email these terms apply to the relevant transaction between us.   
> 
> The contents of this email message and any attachments are intended solely for the addressee(s) and contain CONFIDENTIAL and/or privileged information and are legally protected from disclosure. If you are not the intended recipient of this message, please immediately alert the sender by reply email and then delete this message including any attachments and you are hereby notified that any use, dissemination, copying, or storage of this message or its attachments is strictly prohibited. 
> 
> Full security of emails over the internet cannot be ensured. Despite our efforts it is your responsibility to provide for your protection.
> 
> Global HQ: THEO Technologies NV, Philipssite 5 bus 1, 3001 Heverlee, Belgium - BE 0847.829.290 - CEO: Steven Tielemans
> 
>  
> 
>  
> 
> -- 
> Hls-interest mailing list
> [e-mail] <mailto:[e-mail]>
> https://www.ietf.org/mailman/listinfo/hls-interest <https://www.ietf.org/mailman/listinfo/hls-interest>
> -- 
> Hls-interest mailing list
> [e-mail] <mailto:[e-mail]>
> https://www.ietf.org/mailman/listinfo/hls-interest <https://www.ietf.org/mailman/listinfo/hls-interest>
> 
> 
> -- 
> 
> [ David Hassoun | CEO ] 
> [ p. 303.872.0442 | c. 303.359.7466 ]
> [ RealEyes | www.realeyes.com <http://www.realeyes.com/> ]
> [ 940 Logan | Denver CO, 80203 ]
> 
>  <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=icon>	Virus-free. www.avast.com <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=link> <x-msg://11/#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>-- 
> Hls-interest mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/hls-interest