Implementation Recommendations
Cluster
It is RECOMMENDED that each individual Cluster element contain no more than
five seconds or five megabytes of content.
SeekHead
It is RECOMMENDED that the first SeekHead element be followed by a Void element to
allow for the SeekHead element to be expanded to cover new Top-Level Elements
that could be added to the Matroska file, such as Tags, Chapters, and Attachments elements.
The size of this Void element should be adjusted depending on the Tags,
Chapters, and Attachments elements in the Matroska file.
Optimum Layouts
While there can be Top-Level Elements in any order, some orderings of elements are better than others.
The following subsections detail optimum layouts for different use cases.
Optimum Layout for a Muxer
This is the basic layout muxers should be using for an efficient playback experience:
SeekHeadInfoTracksChaptersAttachmentsTagsClustersCues
Optimum Layout after Editing Tags
When tags from the previous layout need to be extended, they are moved to the end with the extra information. The location where the old tags were located is voided.
SeekHeadInfoTracksChaptersAttachmentsVoidClustersCuesTags
Optimum Layout with Cues at the Front
Cues are usually a big chunk of data referencing a lot of locations in the file.
Players that want to seek in the file need to seek to the end of the file
to access these locations. It is often better if they are placed early in the file.
On the other hand, that means players that don’t intend to seek will have to read/skip
these data no matter what.
Because the Cues reference locations further in the file, it’s often complicated to
allocate the proper space for that element before all the locations are known.
Therefore, this layout is rarely used:
SeekHeadInfoTracksChaptersAttachmentsTagsCuesClusters
Optimum Layout for Livestreaming
In livestreaming ((#livestreaming)), only a few elements make sense. For example, SeekHead and Cues are useless.
All elements other than the Clusters MUST be placed before the Clusters.
InfoTracksAttachments(rare)TagsClusters