MCF vs Matroska

[Eric B]

In year 2002, lots of people talked about Media Container Format (MCF). Any news?
Since first alpha release of matroska, we are all testing this new contenair.
I'm curious about the differences between the 2 projects...

[Atamido]

MCF is very similar to MKV. The main difference is that the fields in MCF are a fixed size. This makes it very easy to parse the file and detect corruption. Matroska was broken off from this project in late 2002 because one of the head programmers (robUx4) wanted to make all of the fields variable size, but the project leader was opposed. The reason to make the fields variable sized was to increase the flexibility of the format. Unfortunately making variable sized fields increases the complexity of the format, and requires more CPU to parse it. Fortunately this overhead turned out to be so small that you will likely never notice it.

Currently the MCF project leader is currently busy/gone, so there has been no progress on it since the split. But most of the MCF team moved to Matroksa, so the idea started by MCF is still living through Matroska.

[Tuning]

Sorry for bringing up this thread.
Is there more difference than quoted in above post?.(I think there is..)
What is the real advantage of MCF?

[Atamido]

That is pretty much it. A file structure using fixed size fields like MCF would be easier to parse for a hardware device. But most hardware devices that were being aimed at are ones that support MPEG-4. These SHOULD have enough to parse Matroska files fine.

[bergi]

But most hardware devices that were being aimed at are ones that support MPEG-4. These SHOULD have enough to parse Matroska files fine.

You are right, but i think it's the not the right level. For example HTTP, you can gzip the answer file, but not the header.
I'm not sure, but i think matroska has also crc, but why crc if you can't even find the header?
Matroska is alright for PC users, but i don't think it has future in hardware devices.

[robUx4]

The CRC verifies the levels below. And all Level0/Level1 elements have an ID (syncword if you prefer this term) or 4 bytes. That should be enough to distinguish it from noise or errors. And this was done with networking/broadcasting in mind.