Welcome to ExtractStream

Faye Pearson faye at c...
Thu, 6 Sep 2001 10:44:56 +0100


From: nickhull@y...
> I've been spending more time looking at the TiVo in the past week. 
> The Summer months have been too busy for me to provide any input. I 
> am impressed with the recent developments and pleased to see that 
> someone has written a PC based decoder (ConvertStream).

The man himself :)

I only made a few meagre changes to ExtractStream and it seems that
with the latest CVS, even fewer changes are needed. It will extract
a 1.5.2 stream just fine when compiled to run on the TiVo (something
I got to test yesterday when my TiVoNET arrived in the post.

> If anyone is interested - that was how ExtractStream was partly 
> developed. I am not a Linux programmer, purely writing for Windows 
> PC's. All the initial work was done in Windows using raw dumps from a 
> donor's TiVo for data. Someone else in the team took my code and made 
> a TiVo version for testing, then enhanced it with convoluted 
> buffering.

That's amazing. There's no byte ordering code in there. I had to add
ntohl() around much of the code to make it work on a PC... Why did I do
that? Well because without a TiVoNET the only sensible way to get the
movies off was to rip the drives out of the tivo, mount them in my Linux
box and do a straight write.

> I wrote the initial mplex supplied in the first version. This was 
> basically a hack of the Linux mplex, but converted to do one pass 
> conversions. Previously it required two passes through the mpeg in 
> order to multiplex. I stripped away all but the most essential code, 
> and then used some calculated constants to fill the gaps. If 
> anything, the mplex program was what I took the most pride in. This 
> code allowed an mpeg to be played straight from the TiVo onto a PC.

Ah, mplex. I haven't looked much at that code. I guess that's the
bit that works on a PC :) I've been doing my multiplexing with
TMPGenc.

> One thing that I am surprised about - nobody has made significant 
> changes to the data search algorythms. We knew that they were buggy 
> at the time of release, but little has been done with them. Maybe an 
> ex-member of #tivo would care to enlighten us.

That would be really cool. It would have been great if there were
some comments in that black magic calculation of block from fsid.

-- 
Faye Pearson,
Software Development Manager,
ClaraNET Ltd. Tel 020 7903 3000

SEMPER UBI SUB UBI!!!!