Comment Re:"memory-safe" (Score 1) 35
Query parameters are URL encoded, not base64 encoded. And these aren't a signficant part of any code profile, you're doing it once per transaction for a tiny handful of strings.
Realistically, the amount of time any web service handler (for example) spends in a base64 decoder is a tiny fraction of any processing they're likely to do.
Be real here. If you don't have access to a profiler, the #1 thing to look at is what's going on inside your loops. You almost never have a base64 decoder being called inside a loop. If you do, it's likely because you're the one making the webservice calls, and the time taken to decode a base64 string is still negligible compared to the time taken to set up and tear down (and wait for!) the network services.
Anyone who claims that this optimization helped speed up their code in any noticeable fashion is either using an extraordinary use-case that likely one off, or is just plain lying. And that's assuming it even, say, doubles the speed of a base64 decode - the GGGP claimed to have written one in pure modern Java and "It matches native speeds within a few %".
What stopped Oracle from doing that? Why would that not be acceptable?
Stop defending mediocrity. Oracle has no idea what it's doing at this stage and is breaking things for no good reason. And it's especially bad because nobody uses Java for speed, they use it for safety, so increasing the speed of something that doesn't need to be sped up, and doing it in a really risky way, is so dumb it's hard to imagine the mentality behind it.