What are the differences between 'call-template' and 'apply-templates' in XSL?

<xsl:call-template> is a close equivalent to calling a function in a traditional programming language.

You can define functions in XSLT, like this simple one that outputs a string.

<xsl:template name="dosomething">
  <xsl:text>A function that does something</xsl:text>
</xsl:template>

This function can be called via <xsl:call-template name="dosomething">.

<xsl:apply-templates> is a little different and in it is the real power of XSLT: It takes any number of XML nodes (whatever you define in the select attribute), processes each of them (not necessarily in any predefined order), somebody could say that apply-templates works like a loop, but this is not exactly the case, as the nodes may be processed in any order, even in parallel, and finds matching templates for them:

<!-- sample XML snippet -->
<xml>
  <foo /><bar /><baz />
</xml>

<!-- sample XSLT snippet -->
<xsl:template match="xml">
  <xsl:apply-templates select="*" /> <!-- three nodes selected here -->
</xsl:template>

<xsl:template match="foo"> <!-- will be called once -->
  <xsl:text>foo element encountered</xsl:text>
</xsl:template>

<xsl:template match="*"> <!-- will be called twice -->
  <xsl:text>other element countered</xsl:text>
</xsl:template>

This way you give up a little control to the XSLT processor - not you decide where the program flow goes, but the processor does by finding the most appropriate match for the node it's currently processing.

If multiple templates can match a node, the one with the more specific match expression wins. If more than one matching template with the same specificity exist, the one declared last wins.

You can concentrate more on developing templates and need less time to do "plumbing". Your programs will become more powerful and modularized, less deeply nested and faster (as XSLT processors are optimized for template matching).

A concept to understand with XSLT is that of the "current node". With <xsl:apply-templates> the current node moves on with every iteration, whereas <xsl:call-template> does not change the current node. I.e. the . within a called template refers to the same node as the . in the calling template. This is not the case with apply-templates.

This is the basic difference. There are some other aspects of templates that affect their behavior: Their mode and priority, the fact that templates can have both a name and a match. It also has an impact whether the template has been imported (<xsl:import>) or not. These are advanced uses and you can deal with them when you get there.

To add to the good answer by @Tomalak:

Here are some unmentioned and important differences:

xsl:apply-templates is much richer and deeper than xsl:call-templates and even from xsl:for-each, simply because we don't know what code will be applied on the nodes of the selection -- in the general case this code will be different for different nodes of the node-list.
The code that will be applied can be written way after the xsl:apply templates was written and by people that do not know the original author.

The FXSL library's implementation of higher-order functions (HOF) in XSLT wouldn't be possible if XSLT didn't have the <xsl:apply-templates> instruction.

Summary: Templates and the <xsl:apply-templates> instruction is how XSLT implements and deals with polymorphism.

Reference: See this whole thread: http://www.biglist.com/lists/lists.mulberrytech.com/xsl-list/archives/200411/msg00546.html

xsl:apply-templates is usually (but not necessarily) used to process all or a subset of children of the current node with all applicable templates. This supports the recursiveness of XSLT application which is matching the (possible) recursiveness of the processed XML.

xsl:call-template on the other hand is much more like a normal function call. You execute exactly one (named) template, usually with one or more parameters.

So I use xsl:apply-templates if I want to intercept the processing of an interesting node and (usually) inject something into the output stream. A typical (simplified) example would be

<xsl:template match="foo">
  <bar>
    <xsl:apply-templates/>
  </bar>
</xsl:template>

whereas with xsl:call-template I typically solve problems like adding the text of some subnodes together, transforming select nodesets into text or other nodesets and the like - anything you would write a specialized, reusable function for.

Edit:

As an additional remark to your specific question text:

<xsl:call-template name="nodes"/>

This calls a template which is named 'nodes':

    <xsl:template name="nodes">...</xsl:template>

This is a different semantic than:

<xsl:apply-templates select="nodes"/>

...which applies all templates to all children of your current XML node whose name is 'nodes'.

The functionality is indeed similar (apart from the calling semantics, where call-template requires a name attribute and a corresponding names template).

However, the parser will not execute the same way.

From MSDN: