Chosen, the ones I am talking about are the ones listed here:
http://www.lslwiki.net/lslwiki/wakka.php?wakka=ClientAssetKeys
If you wanna spend the time to prove it to yourself, EVERY SINGLE ONE of those are sitting there on your HD in the C:\Program Files\SecondLife\skins\textures folder. The only exception, as you note, is the default plywood texture. I imagine that it is probably either hardcoded as a resource, or is located elsewhere, but it is still local.
I'm NOT talking about the Linden Library stuff. Of course that isn't already cached; it's a public asset set, nothing more. You still have to download anything in there.
However, you are talking about things like the default texture, the blank texture, and ones like it. Those ARE always cached, because they are always present in the above folder.
Seriously.
http://www.lslwiki.net/lslwiki/wakka.php?wakka=ClientAssetKeys
If you wanna spend the time to prove it to yourself, EVERY SINGLE ONE of those are sitting there on your HD in the C:\Program Files\SecondLife\skins\textures folder. The only exception, as you note, is the default plywood texture. I imagine that it is probably either hardcoded as a resource, or is located elsewhere, but it is still local.
I'm NOT talking about the Linden Library stuff. Of course that isn't already cached; it's a public asset set, nothing more. You still have to download anything in there.
However, you are talking about things like the default texture, the blank texture, and ones like it. Those ARE always cached, because they are always present in the above folder.
Seriously.
Perhaps you should not have said "all Linden assets" then. If you only meant certain ones, wouldn't it have made sense not to use the word "all"?
In any case, as I said, all we're talking about is the exchange of one small drawback in the form of an extra download for an ongoing benefit in the form of a savings in texture memory. So, while from a networking standpoint, it's better to use the standard blank texture, from a local performance standpoint, it's better to use the custom smaller image.
But either way, it's really not worth worrying about. While it may be of some interest for academic discourse among those of us who like to obsess over technical details, it has virtually no value in practice. As I've said several times now, all that's at stake is a maximum of a few bytes worth of download and/or 3.75 kilobytes worth of texture memory. Neither of those things is ever going to amount to any noticeable difference. It's a drop in the ocean.
