In this this series of blog posts titled “Navigating Cloud DVR’s Murky Legal Waters,” we will first cover the copyright aspects of running a cloud DVR service and explain why some TV service operators choose to implement a private copy mode cloud DVR solution. In the second blog post, we will elaborate on the ramifications of copyright and some court rulings on the technical implementation of cloud DVR solutions and possible legally acceptable optimizations.
– Part 2 –
Implementing Private Copy Cloud DVR
In our first blog in the series, we discussed the impact that copyright law has on cloud DVR service providers. Unless the service provider manages to successfully negotiate the carriage rights with ancillary rights covering the cloud DVR use cases, they will need to exploit the private copy use exception typically found in the copyright laws. In this blog post we will discuss the implementation considerations entailed with deploying private copy use.
2 Implementation Considerations
As discussed in the previous blog post, in general, creating copies of legally obtained copyrighted material for personal use is permitted under a country’s copyrights law. Many of these copyright laws also directly or indirectly address the STB-based DVR use case (STB-DVR) within the scope of the private copy exception. This is based on the assessment that with a STB-DVR the following holds:
- The end user of the device takes the action to make the copy (personal recording)
- The recordings are saved in a protected manner locally (personal [protected] storage)
- The recording playback is limited to the subscriber’s home (non-public broadcast/performance)
Given the parallel between the STB-DVR and cloud DVR functionality one would expect that creating a cloud DVR implementation that exploited private copy exception would be straightforward. However, we’ll examine why things are not so clear-cut.
The most straightforward way to implement a cloud DVR service is to create it in a shared copy mode. Technically this means that only one single recording will be made of a program, irrespective of whether one or more subscribers requested it.
In case of a shared copy cloud DVR implementation, the decision to record may be made upfront by the TV service provider, specifically when implementing a catch-up/restart TV functionality and using this initial recording as the source of the personal recordings.
Although individual subscribers schedule the recording of a program, the implementation creates only one copy from the ingested channel and stores it locally on the TV service provider’s infrastructure. This copy is subsequently used as the master source for playing back the event to subscribers at a later moment in time.
From a copyright perspective such an implementation has been legally challenged on the grounds of right to copy and the right to perform criteria applied to the private copy exception.
Firstly, because it is the TV service provider creating the recordings and doing the so-called illegal copying.
Secondly, given that there is a master copy used to serve all subscriber playbacks we are in fact doing an equivalent to a broadcast, which is beyond the intended use of the private copy exception.
That is unless the country where the TV service provider operates has a specific provision in its copyright law. For example, in Switzerland temporary caching and non-permanent reproduction is allowed, if necessary, as part of communication processes in electronic networks and if they do not have any independent economic significance.
Consequently, TV service providers in Switzerland can legally operate a shared copy cloud DVR service, without explicit carriage agreements with ancillary rights that specifically cover the shared copy mode of cloud DVR.
This has forced some TV service providers to consider implementing cloud DVR in a private copy mode. In such a mode the subscriber determines the start of the recording, and for each individual subscriber request an individual recording is created and stored. Subsequently, this individual copy is used to play back the recording.
This mode, in fact, matches with the way a STB-DVR operates with the only difference being that the storage is no longer local to the STB. Instead, storage is done remotely at the TV service provider’s infrastructure. Consequently, private copy mode is also called remote storage DVR (RS-DVR). Given the resemblance to STB-DVR, operators expected that the private copy exception would be applicable to private copy mode cloud DVR implementations too.
However, copyright owners didn’t see it the same way. Therefore, we have witnessed worldwide different lawsuits to determine the legality of private copy mode cloud DVR services.
In this blog we cannot cover all the cases but let’s review the one that is typically referenced in such litigations, namely the case of Cartoon Network against Cablevision in the U.S.
- The Cablevision RS-DVR case
One of the early court cases was the case of Cartoon Network and a group of other rightsholders against Cablevision in the U.S.
Here the courts had to assess if Cablevision, with their remote storage digital video recorder (RS-DVR) a specific Private Copy Mode Cloud DVR implementation was violating the USA Copyright Act of 1976, which confers upon copyright owners the exclusive rights to, among other things, “reproduce the copyrighted work in copies” and “in the case of … audiovisual works, to perform the copyrighted work publicly.”
 Cartoon Network LP, LLP v. CSC Holdings, Inc
 Twentieth Century Fox Film, Universal City Studios, Paramount Pictures, Disney Enterprises, CBS, ABC , NBC Studios, CNN
Cartoon Network alleged that Cablevision, through its cloud DVR service, directly infringed upon its copyrights in two ways:
1) Cablevision makes unauthorized copies of Cartoon Network’ programming, in violation of Cartoon Network’ right to reproduce their work; and
Multiple unauthorized copies of programming are made in two respects:
(1) a complete copy of a program selected for recording is stored indefinitely on the customer’s allotted hard drive space on the server at Cablevision’s facility; and
(2) portions of programming are stored temporarily in buffer memory on Cablevision’s servers before being transferred to the customer’s allotted hard drive.
2) Cablevision makes unauthorized transmissions of Cartoon Network’ programming, in violation of Cartoon Network’ exclusive right to publicly perform their work.
Although initially the U.S. District Court for the Southern District of New York ruled in favor of the copyright holders, the U.S. Court of Appeals on Aug. 4, 2008, sided with Cablevision reversing the district court’s decision.
The court concluded that:
(1) fleeting buffer copies (lasting less than 1.2 seconds) were not “fixed” sufficiently and thus not “copies” under the Copyright Act;
(2) the consumers who press record are the ones who are making copies not Cablevision; and
(3) the playback of those copies (each of which was recorded separately for each consumer even if thousands of consumers separately chose to record the same show) was not a public performance since each copy could be played only by the consumer who recorded it.
- Legal consequences
Although we have not covered all court cases and their final rulings, there are several legal subtilities to consider when deploying a cloud DVR solution in private copy mode.
However, the different legal court cases that challenged private copy cloud DVR have either explicitly or implicitly defined some additional boundaries on the way to implement such a solution.
Let’s look at the Cablevision RS-DVR case in the U.S. as an example. Because in the court proceedings the legality of the use of a “fleeting buffer” was assessed, they implicitly defined the “legally safe” size of such a common ingest pipeline buffer. In this case it was 1.2 seconds. In other words, if you stay below 1.2 second you cannot be legally challenged; however, if you would have more than 1.2 seconds of common processing buffer you are potentially exposed to litigations.
By the same token, the practice of creating a private copy per subscriber has been deemed acceptable; however, the court has not assessed any technical optimizations who for example from the outside appears to retain the individual copy instances, but from a storage perspective reduces the effective storage requirements.
Equally, the cases discussed the legality of the playback and the requirement that the playback needed to be per subscriber and not shared, as not to become a “public performance.”
However, here we also have several technical optimization options, who ensure that from the outside the requirements are met, but where the [internal] implementation is much more cost-effective.
This means that as a cloud DVR service operator you have some technical freedom to optimize the private copy implementation, but from a pure legal perspective you are not 100% protected.
Hence it is important to work with a cloud DVR solution that offers you the broadest technical optimization options so that at launch you can start with a mode that your legal counsel and the business owners are most comfortable with. Then you’ll have the flexibility to switch to another approach in the case of unfavorable litigation outcomes.
We should mention that quite a few countries are updating their copyright law to create a better legal framework for cloud DVR services. Equally, as copyright owners and broadcasters over time get more familiar with the possibilities enabled by cloud DVR and associated monetization capabilities like dynamic ad insertion, they become more receptive to agreeing with carriage agreements covering cloud DVR use cases in shared copy mode. The selected solution should also have the capabilities to switch from private copy mode to shared copy mode at the TV channel level or if needed at the individual program level.
3 Technical ramifications
This section will discuss the technical implications of the specific legal context and the impact of some legal precedents.
For this we will assess the implementation of the three main functions of a cloud DVR solution: The ingest, playback, and storage processes.
- Shared Copy Mode
As discussed, if allowed to operate in a shared copy mode, it is possible to use an implementation where a single process per TV channel is used to ingest the TV channel. If either the carriage agreement enables it, or it is stipulated as a legal use case, ingested content can be used to create a catch-up TV/restart TV service typically using a circular buffer with a duration of a couple of days. This circular buffer can be used to implement a personal recording service, where individual subscribers select individual programs or program windows/durations to be retained and played back beyond the catch-up TV window. This is possible by copying the matching time range from the shared circular buffer to a shared copy version of the program.
- Private Copy Mode
Taking into account the legal precedents in the U.S. as an example, then we need to also consider the stipulations around the fleeting buffer.
As discussed earlier the court agreed that the 1.2 seconds used in the RS-DVR solution of Cablevision was small enough. If you do not want to take a risk as it relates to the ingest, it means that you will need to implement an ingest process that at no moment in time holds more than 1.2 seconds of video channel in a shared buffer.
In practice this translates into having to implement a method that does ingest processing on a per-subscriber basis. In other words, if you have 1000 subscribers who requested the recording of the same program you will need to have 1000 individual ingest processes (generally under the form of threads) running in order to process the recording of that single program.
It should be noted that given that adaptive bitrate (ABR) streaming is tables stakes for every cloud DVR service, we need solutions capable of ingesting multiple renditions within the same constraints.
However, as these streams are in essence sequences of segments of the video, we tend to see implementations where the ingest timing is synchronized with the segment duration. With segment durations ranging between two seconds and possibly 10 seconds we inevitably end up with ingest buffers larger than 1.2 seconds.
That said, it might be that in other countries the court rulings were not so explicit as it relates to the ingest buffer duration and therefore a substantially more relaxed approach can be taken. More specifically, you could leverage a large shared ingest window — covering the expected largest program duration (for example five hours) —and use that window as the source for the individualized recordings.
- Shared Copy Mode
If allowed to operate in shared copy mode, the playback process is straightforward as the circular buffer can be used in case of a catch-up TV/restart TV playback request. In the case of a personal recording playback request the Content Management System (CMS) will need to determine if the program was already copied from the circular buffer or not. In the latter case this playback is handled as a regular catch-up TV playback, in the former case the shared recording can be used as the source for the playback. Given the shared nature, the resulting video streams can be effectively cached by the operator’s CDN, further enhancing the playback experience for subscribers requesting the same piece of recording.
- Private Copy Mode
Most of the court cases assessed the legality of the private copy cloud DVR implementation based on whether the recordings were subsequently “publicly” performed or not. Here the conclusion was that given that the person who performed the recording is also requesting the playback and that the playback is limited to the household, it was deemed “fair usage” and legally acceptable.
These cases never discussed whether one had to use the original/individualized recording or not for such playback.
Therefore, this opens the possibility to leverage some additional technical optimizations related to the playback (and storage). To ensure that the playback is individual it’s important to ensure at the minimum that the playback requests are unique per subscriber and recording.
Within this approach there are two possible sub-approaches:
- The returned manifest references per subscriber and recording unique segments (the private copy).
- Or it references shared segments based off the master instance of the recording.
In the latter case, we can potentially leverage a CDN to further scale the delivery.
Using only a master instance for playback, the storage can be further optimized, as discussed in the next section.
- Shared Copy Mode
When allowed to operate in shared copy mode service providers are not burdened with having to store a recording of a program for each individual subscriber who made a recording request. But nevertheless, as the cloud DVR service adoption by subscribers increases, TV service providers can expect the storage volume to continue to increase. This is typically addressed by implementing a storage quota per subscriber expressed as a number of hours. Within this quota recordings need to be manually deleted by the subscriber, or alternatively they can be automatically deleted if they are older than a certain age and not explicitly marked for retention by the subscriber.
- Private Copy Mode
When forced to operate in private copy mode, service providers immediately face the challenge of dealing with the massive amount of storage that is required to store all the duplicate copies of the recordings. One way to maintain some control over the storage is to implement a storage quota per subscriber. However, within the private copy mode there are further storage optimization options available.
As most cloud DVR services have support for ABR streaming, there is a requirement to store private copy instances for all the video qualities being used.
However, rather than storing all the video qualities for all the subscribers, service providers have the option of retaining all the qualities for only the first subscribers’ recording (master recording instance). For all other subscribers who requested the recording of the same program, service providers can retain only the lowest quality layer. This can be done immediately at the moment of recording, or alternatively after some period of time. In case of such a storage optimization, service providers will always have — for audit purposes —a private copy of the recording available for each subscriber, albeit only at the lowest quality. For playback, the master recording instance with all the video qualities for all subsequent playback requests can be used.
If the cloud DVR service offers recordings driven via an EPG, there is the option to only retain the private copy instance of the first subscriber as a true physical file and use symbolic links to that file for all other subscribers. This level of optimization can be leveraged either immediately at time of recording, or after some time. In the case of an audit, the cloud DVR service provider can present unique instances for each of the subscribers’ recordings; however, from a storage perspective they will only be taking up the space of a single recording. From a playback perspective, the symbolic links will be used in the underlaying implementation and consequently using the so-called individual subscriber’s instance when playing back in such a scenario.
If the file-based de-duplication approach is deemed too legally risky, there is also the possibility to use a storage subsystem that implements block-level deduplication. They divide each file into smaller blocks and checksum each block individually. Blocks with the same checksum are stored only once. The advantage of this method is that in case, for example, a rerun of a program is recorded it can be mapped at the block level onto the blocks of the initial program airing recording, resulting in effectively better storage optimizations. However, calculating and maintaining the checksums for large number of blocks can be resource intensive, in terms of memory and processor usage.
High-Profile Storage AND Just-In-Time Transcoding (JITT)
As a variation to the low-profile approach, it is also possible to store the highest-quality video layer of all recordings and rather than leverage the video quality ladder of the first subscriber’s recording, run a just-in-time transcoding (JITT) process against the individual’s high-quality recording in order to generate the required intermediate video qualities on the fly. This approach is a trade-off of storage costs versus transcoding costs, but is the least legally challengeable, as each subscriber has their own copy, and the playback is happening against that individual copy.
It’s important note that this is also an optimization technique that can be combined with a shared copy mode, especially for longer tail recordings.
In case we leverage a master copy for the playback, we only need to retain the individual recordings for audit purposes. Therefore, there is the possibility to store individual recordings to a storage type, which is cheaper and/or operationally more efficient. For example, service providers can leverage storage clusters with hibernating disks or even tape robots, which typically come with substantially slower retrieval capabilities.
4 Financial Implications
As discussed in the previous section there are quite a few technical options to implement a cloud DVR solution, ranging from the most cost-effective Shared Copy Mode through the costly 100% Private Copy Mode.
However, as several TV service providers have already demonstrated there is possibility to successfully launch and run cloud DVR services which further optimize the Private Copy mode, especially in terms of storage optimizations. As can be seen in the included graph, which shows normalized cost of the most commonly used implementation options, these optimizations represent substantial cost-savings and consequently can make a material difference in a Cloud DVR service business case.
In this second blog of our series, we examined the legal framework in which cloud DVR services need to operate if they want to exploit the private copy use exception found in most copyright laws. We discussed the different technological optimizations that can be leveraged for a cloud DVR implementation using shared copy or private copy mode.
However, as these technical optimizations have not yet been challenged in court, their legality is not confirmed. As such, it is important to use a cloud DVR solution that gives you the flexibility to adjust quickly in the event of an unfavorable legal outcome.
To learn how Broadpeak can help you navigate the murky cloud DVR legal waters with its cloud PVR solution, check out our “Video Recording in the Cloud: Use Cases and Implementation” white paper.