Stubbisms – Tony’s Weblog

April 28, 2010

Linus’ Git Talk Index

Filed under: Java — Antony Stubbs @ 2:57 pm

Youtube won’t let me post big enough comments, so I’m putting it here.

12:00 Distribution
15:00 CVS Merge Branching
17:50 Perforce at Google
18:18 Committers – Ostensibly not morons
21:30 Question 1 – how to migrate big code bases
22:00 An actual example of distribution
27:00 Question 2 – Politics have just been shifted? You still have to merge. How?
27:38 Networks of trust and masturbation
28:00 The way Linus and Kernel development works
31:41 Question 3 – Are any companies using distributed systems? There seems to be a risk of code bases diverging too much.
32:30 Mercurial – the only other open source distributed VCS worth looking at. Everything else is either centralised, or too slow or unstable.
32:55 Question 3.5 – But is there an advantage for a company to have a centralised model?
33:00 Companies think there is, but there isn’t.
33:20 Git subversion (svn) interfacing
34:55 Question 4 – How does merging work in git in regards to merge conflicts?
37:55 Question 5 – Why is it necessary to be distributed in a company, if you had a system with really good merging?
38:30 Centralised systems work, but cannot work _as well_ as distributed systems.
40:00 Performance
40:40 Branch visibility
42:40 Ugly and stupid
43:15 Question 6 – How many files can git support?
45:54 Big SVN repositories
47:36 Projects that share code
49:25 Question 7 – If you have a huge tree, can you check out only part of it?
50:20 Performance – it’s not how fast you do it, it’s what it enables you to do differently
51:20 SVN designers are complete morons. It’s not branching – it’s the merging
55:15 Git implementation – simple data structures, complicated source code
56:14 Trust and reliability – data protection and security
58:48 Linus’s guarantee – get out what you put in
1:00:05 Trust and Google’s “just not that trustworthy” code hosting
1:03:00 Content vs files – what changed?
1:05:53 Question 8 – Would switching from Perforce just swap scalability problems for other scalability problems?

Blog at