Indeed, ACLs can be used to block certain things very well, and set up basic rules that can limit a box or boxes from accessing crtical things.
However, IPV4 can be anonymized, spoofed, and many folks use hardware firewalling also.
With Linux as router, or BSD, ACLs are valuable, you can build a router that you can dynamically admin to make a gateway that limits LAN access at the gate.There (IE in that scenario), ACLs make hyper-good sense. So does filling the concommitent need to log ACL actions that are applied, in such an application, partly because with many ACLs you might have rules that are residuals and need to be altered.
Build ACLs based on what the ACLed thing is protecting-- and in most cases that is to protect the protected boxes from system compromizes,takeovers,and unwanted data mining. Use ACLs on a basis of who needs to know, and what IP their legitimate access box or ID can use. Actually, for end node protection of non-server apps and boxes, default install automatic build of basic ACLs on a Linux install of Mandrake 9 or 10, SuSE 9.0 and up, FreeBSD 5.2 and up, you will get port based rules in place that give you a good starting point. Use ACLs to supplement this, not replace this. Block with ACLs mostly at gate, and if you have a hyper diverse needs network as far as info needs, then consider security boxes at the major WAN\LAn gateways to LANs by workgroup (segments) and build a WAN\Lan structure into your growth plans.
Use a mix of all the tolls you can get to block viruses, Anti-spam of trainable and untrainable kind is best first bang-for-buck along with gated segments because even today 90% of the viruses come in through email. And, if you can, get your ISP or bandwidth provision folks to preprocess your email with good AV at mail servers that have AV with at least daily updating of defs and heuristics. That will cut your local needs, and since some folk's spam is wanted by others, for individual workstations, trainable Bayesian plus user dynamic customizable anti-spam is also appropriate on workstations as part of an anti-compromising defense.
Use segments as defense zones, layer them, use ACLS mostly at the gateway, and use port rules as much as possible to minimize admin load and admin employee costs related to ACL maintainance. POrovide feedback and loggin means to troubleshoot and find new things that are hitting, and consider IDS also.
With IPV6 implamented this will become something that can be more fiunely tuned,.and in an all-IPV6 world ACLs will play a more important role in gates, but we are not there yet.
Look at how IPV6 by its nature includes MAC info as part of its structure when considering this, please. IPV4 does not. IPV6 allows for so much segmenting that it might be better to more4 granularly segment and anti-spam more and ACL less once you have folks that understand IPV6 in detail.