From mboxrd@z Thu Jan 1 00:00:00 1970 Message-ID: To: 9fans@cse.psu.edu From: rog@vitanuova.com In-Reply-To: <008e01c344ce$b04576c0$d2944251@insultant.net> MIME-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Content-Transfer-Encoding: 7bit Subject: [9fans] simple, sufficient Date: Tue, 8 Jul 2003 00:23:10 +0100 Topicbox-Message-UUID: ed76257a-eacb-11e9-9e20-41e7f4b1d025 > > sufficient for what? > > sufficient to do the job. everything is a trade-off ... there are many different jobs... some might even require >4GB unit transfer sizes to accomplish the job efficiently enough. for another, maybe 9p2000's extra 8 byte overhead on a Tread compared to old 9p might make a crucial difference. "simple and sufficient" is a platitude: i don't think anyone really wants to introduce complexity for its own sake, and what's the point in implementing something if it's not sufficient for the task you want to solve? what i think we're really after is... simple and *insufficient*; :-) choose the things you *won't* be able to do (e.g. read >4GB in a chunk, associate arbitrary metadata on a file) in order to gain simplicity overall. isn't this the reason for plan 9's beauty, and conversely perhaps, one reason for its lack of popularity? trade-offs (particularly trade-offs involving the sacrifice of some potential feature in return for ill-defined simplicity) are hard. many people are unwilling to make them; hence endlessly increasing bloat.