I found a passage in Gildersleeve's grammar on this point. Obviously grammars are fallible but the textual evidence is encouraging:<br /><br />355. The agent of the Gerund and Gerundive is put in the Dative, at all periods.<br />[face=SPIonic][size=18=9]<br /> Diligentia praecipue colenda est nobis, C., Or., II. 35, 148; carefulness is to be cultivated by us first and foremost.<br /><br /> REMARK. -- To avoid ambiguity, especially when the verb itself takes the Dat., the Abl. with ab (a) is employed for the sake of clearness:<br /> Civibus a vobis consulendum, C., Imp., 2, 6; the interest of the citizens must be consulted by you. Supplicatio ab eo decernenda non fuit, C., Ph., XIV. 4, 11<br /><br /> Where there is no ambiguity there is no need of ab:<br /> Linguae moderandum est mihi., Pl., Curc., 486; I must put bounds to my tongue.<br />[/face][/size]<br /><br />
<br /><br />I suppose Gildersleeve is right to point out that not every two uses of a dative with a passive periphrastic are ambiguous. Even though there may be a formal ambiguity among cases, sentences can be constructed so that the meaning is obvious. <br /><br />One could also convey the same meaning in different words: debeo tibi recitare hoc librum, for example. A periphrasis of a periphrasis sounds more painful than it is. Of course a good orator would have handfuls of additional ways to express the thought "I need to" and someone like Cicero will have them all scanned. One could try a construction like "opus est..." or substitute egeo for debeo.