You don't need builtin support of pcontext() functions to use Gettext contexts.

1/ Define the functions yourself:

	function pgettext($context, $message)
	{
		return gettext($actual_message);
	}

	function dpgettext($domain, $context, $message)
	{
		$actual_message = $context . "\04" . $message;
		return dgettext($domain, $actual_message);
	}

	function dcpgettext($domain, $context, $message, $category)
	{
		$actual_message = $context . "\04" . $message;
		return dcgettext($domain, $actual_message, $category);
	}

	function npgettext($context, $msgid1, $msgid2, $n)
	{
		$actual_msgid1 = $context . "\04" . $msgid1;
		$actual_msgid2 = $context . "\04" . $msgid2;
		return ngettext($actual_msgid1, $actual_msgid2, $n);
	}

	function dnpgettext($domain, $context, $msgid1, $msgid2, $n)
	{
		$actual_msgid1 = $context . "\04" . $msgid1;
		$actual_msgid2 = $context . "\04" . $msgid2;
		return dngettext($domain, $actual_msgid1, $actual_msgid2, $n);
	}

	function dcnpgettext(
		$domain, $context, $msgid1, $msgid2, $n, $category)
	{
		$actual_msgid1 = $context . "\04" . $msgid1;
		$actual_msgid2 = $context . "\04" . $msgid2;
		return dcngettext(
			$domain, $actual_msgid1, $actual_msgid2, $n, $category);
	}

2/ Pass additional keywords arguments to xgettext:

	xgettext \
		-kpgettext:1c,2 \
		-kdpgettext:2c,3 \
		-kdcpgettext:2c,3 \
		-knpgettext:1c,2,3 \
		-kdnpgettext:2c,3,4 \
		-kdcnpgettext:2c,3,4

The problem with user implementations is that when a translation isn't found, the cryptic message is returned (the union of context and message). And this why in gettext you use the message as identifier (among other things).

To give full support to gettext, pgettext variant must be implemented, and is an important flag if not implemented, why gettext implemented it if not?

Also, I think that n.oxyde definition of

	function pgettext($context, $message)
	{
		return gettext($actual_message);
	}

Must really be:


	function pgettext($context, $message)
	{
		$actual_message = $context . "\04" . $message;
		return gettext($actual_message);
	}

> The problem with user implementations is that when a translation
> isn't found, the cryptic message is returned (the union of context
> and message).

This is why you should compare the return value of gettext() and
friends in the pgettext() wrappers.  If it is identical to the
context, just return the message ID.  That is exactly how
pgettext() and friends are implemented in GNU libintl.

I don't see the need to implement these functions in ext/gettext.