Обсуждение: on SGML files is used for what ?
Yesterday, Michael Paquier submitted this correction:
Revert "Change the default value of default_toast_compression to "lz4""
which is in the installation.sgml has:
Revert "Change the default value of default_toast_compression to "lz4""
which is in the installation.sgml has:
Build with <productname>LZ4</productname> compression support.
But other SGML files on that commit are all using literal tag.
So I thought, LZ4 is not a product name, so where else is this tag used ?
And that tag is used for Kerberos, Zstandard, GSSAPI, SSPI, Windows, Oracle and others. So, there are company names, programs, protocols, utilities and others.
Do these need to use the <literal> tag, or do we need an additional one, or perhaps a tag for each of these types ?
Marcos
> On 5 Mar 2026, at 19:53, Marcos Pegoraro <marcos@f10.com.br> wrote: > > Yesterday, Michael Paquier submitted this correction: > Revert "Change the default value of default_toast_compression to "lz4"" > which is in the installation.sgml has: > Build with <productname>LZ4</productname> compression support. > But other SGML files on that commit are all using literal tag. > So I thought, LZ4 is not a product name, so where else is this tag used ? > And that tag is used for Kerberos, Zstandard, GSSAPI, SSPI, Windows, Oracle and others. So, there are company names, programs,protocols, utilities and others. The DocBook manual is the main source for truth about when to use the various tags, <productname> is documented in [1]. So Oracle is clearly a productname if it refers to the database and not the company, but Windows should probably use <systemitem class="osname"> [1] instead. Whether or not LZ4 is a product name feels harder to define, but it doesn't seem entirely wrong, especially not for the installation docs. Mind you, we aren't using the tagging exactly according to the specification so appropriate levels of salt should be applied, for example <application> is meant to be used for larger software packages whereas we apply it pg_dump etc. > Do these need to use the <literal> tag, or do we need an additional one, or perhaps a tag for each of these types ? <literal> is IMHO great for tagging text to indicate that it is expected system input/output or source code. Depending on context, lz4 could absolutely use <literal> (but personally I think we should be careful to not overuse <literal>). -- Daniel Gustafsson [0] https://tdg.docbook.org/tdg/5.0/productname.html [1] https://tdg.docbook.org/tdg/5.0/systemitem.html
Em qui., 5 de mar. de 2026 às 19:12, Daniel Gustafsson <daniel@yesql.se> escreveu:
Mind you, we aren't using the tagging exactly according to the specification so
appropriate levels of salt should be applied, for example <application> is
meant to be used for larger software packages whereas we apply it pg_dump etc.
That's correct, but which ones should use different tags, and which tag would be the correct one, because many of them certainly aren't applications.
Should we differentiate them: applications, extensions, languages, protocols, trade names, libraries, ... ?This is the complete list of items used by this tag:
PHP,AIX,Apache,Berkeley Postgres,Bison,Bonjour,Bucardo,bzip2,check_pgsql,
check_postgres,Continuent Tungsten,CrackLib,curl,Curl,Cygwin,Debian GNU/Linux,Diff,DocBook,
DRBD,Emacs,Flex,FOP,FreeBSD,GCC,GDB,Gettext,Git,GNU,GNU Gettext,GNU grep,GNU Libtool,GSSAPI,
gzip,IBM DB2,ICU,ICU4C,Informix,Ingres,Ispell,Kerberos,libc,libnuma,LibreSSL,libselinux,libxml2,
libxslt,Linux,Linux-PAM,LLVM,Londiste,LZ4,macOS,Magicsplat Tcl,Meson,Microsoft Windows,
Microsoft Windows SDK,minidumps,MIT Kerberos,MySQL,NetBSD,NSS,NTLM,nXML Mode,OpenBSD,OpenLDAP,
OpenOffice,OpenSSL,Oracle,ossp-uuid,perf,Perl,pgBadger,pglogical,pkg-config,PL/Proxy,PostGIS,
Postgres,POSTGRES,Postgres95,PostgreSQL,POSTGRES,Version 4.2,Pro*C,Python,Readline,rsyslog,
SELinux,selinux-policy,Slony,Slony-I,Solaris,SSPI,Strawberry Perl,systemd,Tcl,vi,vim,
Visual Studio,Visual Studio 2019,Visual Studio 2022,Visual Studio Express,Windows,
Windows Debugger Tools,Windows SDK,Windows SDK 10,WinLDAP,Wireshark,zlib,Zstandard
check_postgres,Continuent Tungsten,CrackLib,curl,Curl,Cygwin,Debian GNU/Linux,Diff,DocBook,
DRBD,Emacs,Flex,FOP,FreeBSD,GCC,GDB,Gettext,Git,GNU,GNU Gettext,GNU grep,GNU Libtool,GSSAPI,
gzip,IBM DB2,ICU,ICU4C,Informix,Ingres,Ispell,Kerberos,libc,libnuma,LibreSSL,libselinux,libxml2,
libxslt,Linux,Linux-PAM,LLVM,Londiste,LZ4,macOS,Magicsplat Tcl,Meson,Microsoft Windows,
Microsoft Windows SDK,minidumps,MIT Kerberos,MySQL,NetBSD,NSS,NTLM,nXML Mode,OpenBSD,OpenLDAP,
OpenOffice,OpenSSL,Oracle,ossp-uuid,perf,Perl,pgBadger,pglogical,pkg-config,PL/Proxy,PostGIS,
Postgres,POSTGRES,Postgres95,PostgreSQL,POSTGRES,Version 4.2,Pro*C,Python,Readline,rsyslog,
SELinux,selinux-policy,Slony,Slony-I,Solaris,SSPI,Strawberry Perl,systemd,Tcl,vi,vim,
Visual Studio,Visual Studio 2019,Visual Studio 2022,Visual Studio Express,Windows,
Windows Debugger Tools,Windows SDK,Windows SDK 10,WinLDAP,Wireshark,zlib,Zstandard
regards
Marcos
On Fri, Mar 6, 2026 at 10:24:39AM -0300, Marcos Pegoraro wrote: > Em qui., 5 de mar. de 2026 às 19:12, Daniel Gustafsson <daniel@yesql.se> > escreveu: > > Mind you, we aren't using the tagging exactly according to the > specification so > appropriate levels of salt should be applied, for example <application> is > meant to be used for larger software packages whereas we apply it pg_dump > etc. > > > That's correct, but which ones should use different tags, and which tag would > be the correct one, because many of them certainly aren't applications. > Should we differentiate them: applications, extensions, languages, protocols, > trade names, libraries, ... ? > This is the complete list of items used by this tag: > PHP,AIX,Apache,Berkeley Postgres,Bison,Bonjour,Bucardo,bzip2,check_pgsql, > check_postgres,Continuent Tungsten,CrackLib,curl,Curl,Cygwin,Debian GNU/ > Linux,Diff,DocBook, > DRBD,Emacs,Flex,FOP,FreeBSD,GCC,GDB,Gettext,Git,GNU,GNU Gettext,GNU grep,GNU > Libtool,GSSAPI, > gzip,IBM > DB2,ICU,ICU4C,Informix,Ingres,Ispell,Kerberos,libc,libnuma,LibreSSL,libselinux,libxml2, > libxslt,Linux,Linux-PAM,LLVM,Londiste,LZ4,macOS,Magicsplat Tcl,Meson,Microsoft > Windows, > Microsoft Windows SDK,minidumps,MIT Kerberos,MySQL,NetBSD,NSS,NTLM,nXML > Mode,OpenBSD,OpenLDAP, > OpenOffice,OpenSSL,Oracle,ossp-uuid,perf,Perl,pgBadger,pglogical,pkg-config,PL/ > Proxy,PostGIS, > Postgres,POSTGRES,Postgres95,PostgreSQL,POSTGRES,Version > 4.2,Pro*C,Python,Readline,rsyslog, > SELinux,selinux-policy,Slony,Slony-I,Solaris,SSPI,Strawberry > Perl,systemd,Tcl,vi,vim, > Visual Studio,Visual Studio 2019,Visual Studio 2022,Visual Studio > Express,Windows, > Windows Debugger Tools,Windows SDK,Windows SDK > 10,WinLDAP,Wireshark,zlib,Zstandard When to use which tag isn't always clear. When Peter Eisentraut initially did the work, he said just pick the best one, and use <literal> if you are unclear. -- Bruce Momjian <bruce@momjian.us> https://momjian.us EDB https://enterprisedb.com Do not let urgent matters crowd out time for investment in the future.
Em qua., 11 de mar. de 2026 às 14:25, Bruce Momjian <bruce@momjian.us> escreveu:
When to use which tag isn't always clear. When Peter Eisentraut
initially did the work, he said just pick the best one, and use
<literal> if you are unclear.
So, wouldn't it be better to have more labels ? like ...
gzip, zlib, gzip, LZ4 ... <compressionname>
Oracle, MySQL, Ingres, Informix, IBM DB2 ... <databasename>
OpenBSD, AIX, FreeBSD, Windows, ... <osname>
Not a dozen different labels, as it would be difficult to remember which ones are available,
but perhaps 3 or 4 would better detail what each of these items represents.
If someone ever wants to have a different style for one of them, it will be easy to change, right ?
regards
Marcos
On Thu, Mar 12, 2026 at 2:12 PM Marcos Pegoraro <marcos@f10.com.br> wrote: > > Em qua., 11 de mar. de 2026 às 14:25, Bruce Momjian <bruce@momjian.us> escreveu: >> >> When to use which tag isn't always clear. When Peter Eisentraut >> initially did the work, he said just pick the best one, and use >> <literal> if you are unclear. > > > So, wouldn't it be better to have more labels ? like ... > gzip, zlib, gzip, LZ4 ... <compressionname> > Oracle, MySQL, Ingres, Informix, IBM DB2 ... <databasename> > OpenBSD, AIX, FreeBSD, Windows, ... <osname> > > Not a dozen different labels, as it would be difficult to remember which ones are available, > but perhaps 3 or 4 would better detail what each of these items represents. > If someone ever wants to have a different style for one of them, it will be easy to change, right ? > Aside from theoretical correctness, what user problem does this solve? Robert Treat https://xzilla.net
Em qui., 12 de mar. de 2026 às 15:23, Robert Treat <rob@xzilla.net> escreveu:
Aside from theoretical correctness, what user problem does this solve?
Probably nothing for today, but if someone wants to change the style of <productname> or <literal> or <acronym>,
they will also change others that shouldn't be modified.
they will also change others that shouldn't be modified.
And there are some which are wrong, like <application>PostgreSQL</application>
regards
Marcos
On Thu, Mar 12, 2026 at 03:47:01PM -0300, Marcos Pegoraro wrote: > Em qui., 12 de mar. de 2026 às 15:23, Robert Treat <rob@xzilla.net> escreveu: > > Aside from theoretical correctness, what user problem does this solve? > > > Probably nothing for today, but if someone wants to change the style of > <productname> or <literal> or <acronym>, > they will also change others that shouldn't be modified. > > And there are some which are wrong, like <application>PostgreSQL</application> It would be nice to consistently use the same markup for the same word. -- Bruce Momjian <bruce@momjian.us> https://momjian.us EDB https://enterprisedb.com Do not let urgent matters crowd out time for investment in the future.