Received: from sss2.sss.pgh.pa.us (sss.pgh.pa.us [209.114.166.2]) by hub.org (8.10.1/8.10.1) with ESMTP id e681p1W81663 for ; Fri, 7 Jul 2000 21:51:02 -0400 (EDT) Received: from sss2.sss.pgh.pa.us (tgl@localhost [127.0.0.1]) by sss2.sss.pgh.pa.us (8.9.3/8.9.3) with ESMTP id VAA19530; Fri, 7 Jul 2000 21:49:53 -0400 (EDT) To: Bruce Momjian cc: Philip Warner , Jan Wieck , Peter Eisentraut , PostgreSQL HACKERS Subject: Re: Re: [SQL] Re: [GENERAL] lztext and compression ratios... In-reply-to: <200007080119.VAA14375@candle.pha.pa.us> References: <200007080119.VAA14375@candle.pha.pa.us> Comments: In-reply-to Bruce Momjian message dated "Fri, 07 Jul 2000 21:19:50 -0400" Date: Fri, 07 Jul 2000 21:49:53 -0400 Message-ID: <19527.963020993@sss.pgh.pa.us> From: Tom Lane X-Archive-Number: 200007/473 >> This remoinded me of a question I wanted to ask Unix people: other OSs I >> use allow for dynamic linking, at runtime and in code, against shared >> libraries, and I know Unix must allow this. The places where zlib is used >> are pretty limited, so it might be worth considering doing the 'HAVE_ZLIB' >> kinds of checks at runtime. Then one binary fits all... > We do dynamic loading for functions. Not sure if we want to load zlib > dynamically if we can help it. Not bloody likely! Do you want to be in a position where you restart your postmaster and suddenly chunks of your database are inaccessible? That's what could happen to you if someone moves or deletes libz.so. I don't mind being dynamically linked to standard system shared libs; if libc.so is busted then whether Postgres launches is the least of your worries. But dynamic dependence on an optional package that's probably living in /usr/local strikes me as exceedingly risky. If we do go with using zlib instead of homegrown code, I would recommend building and statically linking to our own copy even if there is a copy available on the system. This will prevent cross-version compatibility problems as well as where'd-my-library-go syndrome. We cannot afford those sorts of risks for something that could prevent us from reading our database. regards, tom lane