Michael, this quote from Nevin intrigued, and so I tracked it down.
What I found is an entirely interesting and exceptionally valuable thread—moved to General Discussion of all things by a moderator, and only discoverable in the Archived forum. Thank goodness we still have that.
http://ellislab.com/forums/viewthread/137087/
There are also several linked documents of high value for performance, not all of which I’ve seen before.
In very brief summary, you can find out why add-ons like CE Cache and several of the Solspace ones are out there; not always for reasons you might suspect, as in the case of Super Search.
Further, Paul Burdick, who designed a lot of ExpressionEngine itself and was employed later for SolSpace on performance matters specifically, is out there on Twitter, and I think also still in some connection with Solspace.
On the matter of not caching templates from Nevin, my take is that the issue comes with larger sites, and perhaps especially with embed templates. EE Caching apparently occurs per browser URL, rather than per template as it may seem from its setup—likely it is whether any cache-marked templates are included that determines to cache a given URL.
Thus if you mark an embed, and it is used multiple places, you use up one cache entry for each of those places (urls) towards the limit you’ve set, or the system limit of 1000 cache entries.
On a small site, I’ve measured and found that straightforwardly caching things like embeds gives big speed advantages, and doesn’t run into the limits; thus a simple-minded approach apparently works, which is a bonus for many EE sites.
On a large site…well, read the thread. You will come out educated, and with a great set of information, links, and evident persons, to go farther.
Stunning stuff, which shows what the EE community has been like.
It makes me think there should be an area of the Ellis sites where such gems are very securely collected and visible—and whether they originate in the ‘ordinary’ forums or not, looking ahead to the enterprise support. Also that other policies considered going forward, so that everything deeply informational isn’t bogged down, or transformed into pay-high-to-be-made-aware-at-all.
I think the arguments for this would include that persons need a continuous path to move from self-support towards and into recognizing the time had come for highly paid consultancy; and that certain flavors of working community can generate very surprisingly valuable contributions, which were also not ever expected.
Individuals do good work, and that is after all how the performance teams, CausingEffect, and Solspace (with this hat on at least) got started. That’s on the performance front: think how many other highly appreciated and highly capable products and services have emerged in the rest of the arena, making EE what it now is.
Regards, and off soap box, but it seems an important one,
Clive