Lets kick the vote off now then.... On Sun, 12 Jul 2009 19:51:25 -0600, Alex Rousskov <[email protected]> wrote: > On 07/12/2009 06:28 PM, Amos Jeffries wrote: >> I hate the size of this. I really do. > > I too wish it could be trickled into the trunk in smaller chunks over > time, but I think we should still be grateful that these changes and the > ICAP chaining code got implemented and released. The big-or-nothing > choice is unfortunate, but the trickling overheads are significant and > often exceed our very limited resources. > >> However there are several important 3.1 bugs that is either fixes, or >> adds >> support towards fixing easier. >> >> So, after 3.1.0.10 stabilization update is out I'll be considering the >> 3p1 >> imports for 3.1.0.11. > > I think we should also consider importing the ICAP chains code. Here are > the pros and cons I can think of: > > Cons: > - Violating feature freeze policy sets a bad precedent. > - There is a significant risk of hurting v3.1 stability. > > Pros: > - The changes are often essential for NetCache parity. > - Timing: folks are replacing end-of-life NetCaches _now_. > - Any stability problems will be temporary. > - Changes are easy to import (for now). > - I will waste fewer cycles on 3p1-plus maintenance. > - I may be able to spend more cycles on v3.1 code. > > Squid v3.2 timing is also a consideration, I guess. We do not have a lot > of reliable statistics, but it may take another 10+ months before v3.2 > becomes stable. That version will have enough features to encourage > upgrades even if adaptation chains are already supported in v3.1. On the > other hand, v3.1 can benefit from a feature boost.
Your thinking parallels mine with a few extra Pros I had not thought of. I'm voting +1 to adding both. Amos
